Coordinator's names Carole DELPORTE-GALLET Hugues ...
Coordinator's names Carole DELPORTE-GALLET Hugues ...
Coordinator's names Carole DELPORTE-GALLET Hugues ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Coordinator’s<br />
<strong>names</strong><br />
<strong>Carole</strong> <strong>DELPORTE</strong>-<strong>GALLET</strong><br />
<strong>Hugues</strong> FAUCONNIER<br />
LIAFA-GANG University Paris Diderot<br />
Acronym<br />
DISPLEXITY<br />
Titre de la<br />
proposition du<br />
projet<br />
Calculabilité et complexité en distribué<br />
Proposal title<br />
Distributed Computing : computability and complexity<br />
com-<br />
Evaluation<br />
mittee<br />
SIMI 2- Science informatique et applications<br />
Multidisciplinary<br />
proposal<br />
NO<br />
Type of research<br />
Basic Research<br />
NO<br />
Grant requested 829 515, 47<br />
dura-<br />
Proposal<br />
tion<br />
International cooperation<br />
4 years<br />
1
Contents<br />
1 Proposal Abstract 3<br />
2 Context, Positionning and Objectives of the proposal 4<br />
2.1 Context of the proposal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4<br />
2.2 Objectives, originality and/or novelty of the proposal . . . . . . . . . . . . . . . . . . . . . . . 6<br />
3 Scientific and technical programme, proposal organisation 6<br />
3.1 Scientific programme, proposal structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6<br />
3.2 Description by task . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
3.2.1 Task 1: Project Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8<br />
3.2.2 Task 2: “yes-no” and Decision Problems in Distributed Computing . . . . . . . . . . . 8<br />
3.2.3 Task 3: Oracles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9<br />
3.2.4 Task 4: Complexity classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11<br />
3.2.5 Task 5: Non-determinism in distributed computing . . . . . . . . . . . . . . . . . . . . 12<br />
3.2.6 Task 6: New computational paradigms/frameworks . . . . . . . . . . . . . . . . . . . . 13<br />
3.3 Tasks schedule, deliverables and milestones . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14<br />
4 Dissemination and exploitation of results, intellectuel property 15<br />
5 Consortium description 16<br />
5.1 Partners description and relevance, complementarity . . . . . . . . . . . . . . . . . . . . . . . 16<br />
5.2 Qualification of the proposal coordinator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17<br />
5.3 Qualification and contribution of each partner . . . . . . . . . . . . . . . . . . . . . . . . . . . 18<br />
6 Scientific justification of requested ressources 21<br />
6.1 Partner 1: Paris . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21<br />
6.2 Partner 2: Rennes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22<br />
6.3 Partner 3: Bordeaux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23<br />
7 Annexes 24<br />
7.1 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24<br />
7.2 CV, resume . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26<br />
7.3 Staff involvment in other contracts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44<br />
2
1 Proposal Abstract<br />
Distributed computation keep raising new questions concerning computability and complexity. For instance,<br />
as far as fault-tolerant distributed computing is concerned, impossibility results do not depend on the computational<br />
power of the processes, demonstrating a form of undecidability which is significantly different from<br />
the one encountered in sequential computing. In the same way, as far as network computing is concerned, the<br />
impossibility of solving certain tasks locally does not depend on the computational power of the individual<br />
processes.<br />
The main goal of DISPLEXITY (for DIStributed computing: computability and ComPLEXITY) is to<br />
establish the scientific foundations for building up a consistent theory of computability and complexity for<br />
distributed computing.<br />
One difficulty to be faced by DISPLEXITY is to reconcile the different sub-communities corresponding<br />
to a variety of classes of distributed computing models. The current distributed computing community may<br />
indeed be viewed as two not necessarily disjoint sub-communities, one focusing on the impact of temporal<br />
issues, while the other focusing on the impact of spatial issues. The different working frameworks tackled by<br />
these two communities induce different objectives: computability is the main concern of the former, while<br />
complexity is the main concern of the latter.<br />
Within DISPLEXITY, the reconciliation between the two communities will be achieved by focusing on<br />
the same class of problems, those for which the distributed outputs are interpreted as a single binary output:<br />
yes or no. Those are known as the yes/no-problems. The strength of DISPLEXITY is to gather specialists of<br />
the two main streams of distributed computing. Hence, DISPLEXITY will take advantage of the experience<br />
gained over the last decade by both communities concerning the challenges to be faced when building up a<br />
complexity theory encompassing more than a fragment of the field.<br />
In order to reach its objectives, DISPLEXITY aims at achieving the following tasks:<br />
• Formalizing yes/no-problems (decision problems) in the context of distributed computing. Such problems<br />
are expected to play an analogous role in the field of distributed computing as that played by<br />
decision problems in the context of sequential computing.<br />
• Formalizing decision problems (yes/no-problems) in the context of distributed computing. Such problems<br />
are expected to play an analogous role in the field of distributed computing as that played by<br />
decision problems in the context of sequential computing.<br />
• Revisiting the various explicit (e.g., failure-detectors) or implicit (e.g., a priori information) notions of<br />
oracles used in the context of distributed computing allowing us to express them in terms of decidability/complexity<br />
classes based on oracles.<br />
• Identifying the impact of non-determinism on complexity in distributed computing. In particular,<br />
DISPLEXITY aims at a better understanding of the apparent lack of impact of non-determinism<br />
in the context of fault-tolerant computing, to be contrasted with the apparent huge impact of nondeterminism<br />
in the context of network computing. Also, it is foreseen that non-determinism will enable<br />
the comparison of complexity classes defined in the context of fault-tolerance with complexity classes<br />
defined in the context of network computing.<br />
• Last but not least, DISPLEXITY will focus on new computational paradigms and frameworks, including,<br />
but not limited to distributed quantum computing and algorithmic game theory (e.g., network<br />
formation games).<br />
The project will have to face and solve a number of challenging problems. Hence, we have built the DIS-<br />
PLEXITY consortium so as to coordinate the efforts of those worldwide leaders in Distributed Computing<br />
who are working in our country. A successful execution of the project will result in a tremendous increase<br />
in the current knowledge and understanding of decentralized computing and place us in a unique position<br />
in the field.<br />
3
2 Context, Positionning and Objectives of the proposal<br />
2.1 Context of the proposal<br />
The distributed community can be viewed as the union of two sub-communities. Even though they are not<br />
completely disjoint, they are disjoint enough not to leverage each other’s results. At a high level, one is mostly<br />
interested in timing issues (clock drifts, link delays, crashes, etc.) while the other one is mostly interested in<br />
spatial issues (network structure, memory requirements, etc.). Indeed, one sub-community is mostly focusing<br />
on the combined impact of asynchronism and faults on distributed computation, while the other addresses<br />
the impact of network structural properties on distributed computation. Both communities address various<br />
forms of computational complexities, through the analysis of different concepts. This includes, e.g., failure<br />
detectors and wait-free hierarchy for the former community, and compact labeling schemes and computing<br />
with advice for the latter community.<br />
We illustrate these points by some examples below.<br />
In the wait-free model, each process starts with a private input value, and decides irrevocably a private<br />
output value after a finite number of steps. The processes communicate with each other through shared<br />
objects, but the output decision of each process must occur independently of the speed or actions of other<br />
processes. The output value decided by each process, which must satisfy some input-output specification,<br />
could however depend on the actions of the other processes. A wait-free distributed algorithm performs<br />
correctly in any asynchronous environment, and, in particular, it tolerates an arbitrary number of process<br />
crashes. In this way, the wait-free model may be considered as a special case of t-resiliency for t equal to the<br />
number of processes. Some simulations, like in [4, 5] and general results prove that t-resiliency may generally<br />
be reduced to the wait-free model.<br />
Not all tasks (specified by input-output relations) are wait-free solvable. More specifically, the famous<br />
FLP Theorem [16] states that consensus is not 1-resilient and then not wait-free solvable. Several research<br />
frameworks in the context of wait-free and more generally fault-tolerant computing offer similarities with<br />
computing with oracle machines. One typical example is provided by the failure detector theory [8] which<br />
is based on oracles providing nodes with limited information about the failures. Failure detectors may be<br />
compared by reduction [7]. Hence if every failure detector enabling to solve task A may be reduced to a<br />
failure detector enabling to solve task B, then task A is harder (i.e. needs mode information about failure)<br />
than task B. In this way, failure detectors enable to define hierarchy among unsolvable tasks [10].<br />
Another example is provided by the wait-free hierarchy (including Herlihy’s hierarchy [22]) which maps<br />
object to positive levels such that an object is at level n if and only if it offers some form of universality for<br />
a system of n processes in the wait-free model.<br />
In network computing, networks are modeled by graphs G = (V ; E), in which messages are exchanged<br />
between the processes (i.e., the nodes in V ), along the links of the network (i.e., the edges in E). The model<br />
LOCAL [27] assumes that the nodes operate in synchronous rounds, where a round enables to exchange a<br />
message of arbitrarily large size between every pair of nodes linked by an edge. Hence, in the LOCAL model,<br />
the main measure of interest is the maximum distance at which processes interact. A task (e.g., coloring the<br />
nodes of properly with at most ∆+1 colors) is local if and only if it has an algorithm solving it in O(1) rounds<br />
in the LOCAL model 1 . Not all problems are local. In particular, Linial [25] proved that graph coloring is not<br />
local, even in rings (O(log n) rounds are required to 3-color the ring). Naor and Stockmeyer [26] studied the<br />
class LCL of locally checkable labeling problems. As in the case of wait-free computation, several research<br />
frameworks in the context of local computing offer similarities with computing with oracle machines. One<br />
typical example is distributed computing with advice, which is aiming at capturing the impact of global<br />
structural knowledge (e.g., the size of the network) on the efficiency of solving distributing computing tasks<br />
(like coloring or broadcasting). Another example is proof labeling schemes [24], which is aiming at providing<br />
each node with labels such that the consistency of distributed data-structures (e.g., MST, spanning tree,<br />
etc.) can be checked in one round, just by exchanging labels between neighbors.<br />
1 An interesting restriction do the LOCAL model is the CON GEST (B) model in which the size of exchanged messages is<br />
bounded by B.<br />
4
Although the examples mentioned above offer the flavor of computational complexity, they are not quite<br />
clearly positioned as part of computational complexity. Comparing problems (or, more appropriately, tasks)<br />
that are not solvable in the fault-tolerance models is achieved through the analysis of their weakest failure<br />
detectors, while wait-free reductions are rather performed between objects and not between problems. Similarly,<br />
the concepts of advices and proof labeling schemes enable to handle and sometime compare problems<br />
that are not locally computable. Yet the papers addressing these notions are mostly problem oriented, and<br />
do not provide a global consistent distributed computing complexity framework. It is interesting to note<br />
that although the above examples are dealing with computational complexity, each time, the approach is ad<br />
hoc. We believe that we reach the point where it is necessary to base a general theory of computability and<br />
complexity for distributed computing. Such a theory is crucial to have a better understanding of what is<br />
distributed computing. Considering the relationship between classical algorithmic and classical complexity<br />
theory, it is clear that it can be expected that computability and complexity theory for distributed computing<br />
will have many applications concerning distributed algorithms. The main goal of the project is precisely to<br />
work towards this goal.<br />
This project is clearly very ambitious. Yet, more than 30 years of distributed computing have enabled<br />
to accumulate a very strong knowledge and the partners of this project are at the leading edge of research<br />
on distributed computing and cover a large spectrum of this discipline. Should this project successful, we<br />
foresee a significant impact in the area and a unique position of the French research.<br />
Related ANR project<br />
Some partners (Bordeaux and Paris) participate to the ALADIN ANR Project (2007-2011) that aimed at<br />
studying fundamental aspects of large interaction networks. More specifically, one of the topics of ALADIN<br />
is to deal with the design of distributed algorithms with limited knowledge and/or limited probing capacity.<br />
The ALADIN project will be finished at the end of the year (2011). The works realized by the partners in<br />
this ANR will be helpfull for the project. In particular, some objectives of the DISPLEXITY project comes<br />
from recent results [19, 20] of ALADIN concerning LOCAL distributed decision and decidability classes for<br />
mobile agents computing.<br />
Some partners (Paris and Rennes) participate to the SHAMAN ANR-VERSO (2008-2012) project that<br />
focuses on the algorithmic foundations of resource-constrained autonomous large scale systems. The first objective<br />
of this project is the design of realistic models encompassing anonymity, dynamism, and/or malicious<br />
behavior. This project should give some help concerning some new aspects of distributed computation especially<br />
concerning large scale systems. Many works of Rennes concern oracles (failure detectors)(e.g. [2, 3]).<br />
In [11], some members of Paris and Rennes give a model that encompasses anonymity and malicious behavior<br />
that would be useful for DISPLEXITY and [12] defines the notion of adversaries that will be tackled in<br />
this project in the framework of oracles. On the other hand, the progress on computability and complexity<br />
made by the DISPLEXITY project will help the SHAMAN project to better evaluate the theoretical power<br />
of these models.<br />
Some partners (Bordeaux and Paris) participate to the PROSE ANR Project (Sept 2009-Aug 2012).<br />
The goal is the deployment of social networking applications in a delay tolerant manner using opportunistic<br />
social contacts as in a peer to peer network, as well as new advanced content recommendation engines.<br />
It is a multi-disciplinary project to design opportunistic contact sharing schemes and to characterize the<br />
environmental conditions, the usage constraint, as well as the algorithmic and architecture principles that<br />
let them operate. One task concerns the theory of dynamic graph and networking modeling that is interesting<br />
for DISPLEXITY. In return, PROSE will benefit with help of DISPLEXITY of a better understanding<br />
of this model.<br />
5
2.2 Objectives, originality and/or novelty of the proposal<br />
We have the ambitious project to achieve the reconciliation between the two communities by focusing on<br />
the same class of problems, the yes/no-problems, and establishing the scientific foundations for building up<br />
a consistent theory of computability and complexity for distributed computing.<br />
The main question addressed in this project is the following: is the absence of globally coherent computational<br />
complexity theories covering more than fragments of distributed computing inherent to the field?<br />
One issue is obviously the types of problems located at the core of distributed computing. Tasks like<br />
consensus, leader election, and broadcasting are of very different nature. They are not yes-no problems 2 ,<br />
neither are they minimization problems. Coloring and Minimal Spanning Tree are optimization problems<br />
but we are often more interested in constructing an optimal solution than in verifying the correctness of<br />
a given solution. Still, it makes full sense to analyze the yes-no problems corresponding to checking the<br />
validity of the output of tasks.<br />
Another issue is the power of individual computation. The FLP impossibility result 3 as well as Linial’s<br />
lower bound [25] hold independently from the individual computational power of the involved computing<br />
entities. For instance, the individual power of solving NP-hard problems in constant time would not help<br />
overcoming these limits which are inherent to the fact that computation is distributed.<br />
A third issue is the abundance of models for distributed computing frameworks, from shared memory to<br />
message passing, spanning all kinds of specific network structures (complete graphs, unit-disk graphs, etc.)<br />
and or timing constraints (from complete synchronism to full asynchronism). There are however models,<br />
typically the wait-free model and the LOCAL model, which, though they do not claim to reflect accurately<br />
real distributed computing systems, enable focusing on some core issues.<br />
Despite the above issues, this project seeks to demonstrate that many important notions of Distributed<br />
Computing seem to fit well with standard computational complexity. Distributed Computing should thus<br />
greatly benefit from expressing its main challenges in this standard framework for making them accessible<br />
to a wider audience.<br />
3 Scientific and technical programme, proposal organisation<br />
3.1 Scientific programme, proposal structure<br />
The main purpose of this project is to define the basis for the complexity and the computability of distributed<br />
computing.<br />
First of all, the classical notions of computability in term of, for example Turing machines, do not apply<br />
directly to distributed computing. Distributed computing deals with infinite computations and some models<br />
of computations for distributed protocols (e.g. population protocols [1] or self-stabilizing algorithms [15]) are<br />
defined in terms of convergence (only computations such that eventually the output stabilizes are considered)<br />
and asynchrony in wait-free and fault-tolerant computing are properties on sets of infinite sequences with<br />
some fairness conditions. The FLP impossibility result for the consensus [16] does not depend on the<br />
computability power of the processes, and this result is then very different from the classical undecidability<br />
results with Turing machines.<br />
Concerning complexity problems, distributed computing raises new problems. One of the main point is<br />
here the question of the locality of computations as in the LOCAL [27] or CON GEST models making the<br />
definition of complexity classes relevant for distributed computing necessary.<br />
Nevertheless, even if the classical computability and complexity theory do not apply directly here, it is<br />
clear that most of its methods and tools will give the basis to establish a computational and complexity<br />
theory for distributed computing. Therefore the main approach of this project is to try to use classical tools<br />
2 yes-no problems are generally called decision problems in classical computing theory. They are problems that output yes<br />
or no depending on the inputs of the nodes. A precise definition of yes-no problems in distributed computing is not obvious.<br />
Yes-no problems is the subject of Task 2.<br />
3 FLP result [16] is a fundamental result in fault-tolerant distributed computing that proves the impossibility of consensus<br />
when at least one process may crash.<br />
6
of computability and complexity and to adapt them to the specificity of distributed computing. For, this we<br />
have identified some of them that will serve as central themes of each task.<br />
As for Turing machine, the main goal of Task 2 is to define “yes-no” problems 4 and then to tackle the<br />
relationship between computing and verifying a problem under various computational models for distributed<br />
computing. Note that in the distributed context even the definition of “yes-no” problems is not a priori<br />
obvious depending on who has to decide and several definitions are possible. For distributed computations,<br />
the relationship between computing and verifying is not so obvious. For example with faulty processes it<br />
may be easier to achieve consensus between processes than verifying that consensus is achieved. Definition<br />
and study of “yes-no” problems will enable to tackle this fundamental issue.<br />
Another fundamental tool concerning computability is the notion of oracles, subject of Task 3. Oracles,<br />
like failure-detectors [8] in fault-tolerance computing have been introduced to bypass the impossibility result<br />
of the consensus. Indeed with some partial information about crashes problems like consensus become<br />
possible to solve. Failure detectors encapsulate this information about crashes. They have proved their<br />
efficiency enabling to obtain hierarchies between problems impossible to solve in presence of process crashes.<br />
The oracles are not only useful in fault-tolerant computing to establish hierarchy between (impossible)<br />
problems but, as for classical computability and complexity theory, they are fundamental basic elements of<br />
distributed computing. For example, an oracle that gives the number of nodes enable to distinguish between<br />
rings of nodes that otherwise are indistinguishable. More generally, the oracles and the notion of reduction<br />
with oracles enables to define hierarchy and complexity classes for distributed computing. Hence general<br />
study of oracles and reductions in distributed computing will help to define complexity and computability<br />
classes for distributed computing.<br />
Defining complexity classes for distributed computing comes naturally after the study of “yes-no” problems<br />
and oracles and is the subject of Task 4. Clearly these definitions depend on the definition of the<br />
definitions of “yes-no” problems and oracles. Two main kinds of classes are envisioned: complexity classes<br />
based on the notion of time (or here number of communication rounds) and classes based on the notion of<br />
possibility-impossibility. For example, for the first ones, Local Decision (LD) class is the class of distributed<br />
languages with decision in constant time in the LOCAL model and for the second ones new classes have to<br />
be defined for the wait-free and fault-tolerant models. For these models some complexity classes indexed<br />
by oracles may also be defined. Extension of these classes to probabilistic setting (at least for local decision<br />
classes) may be also considered. The goal is here to define and obtain hierarchies similar to classical<br />
complexity cases.<br />
In classical computability and complexity, non-determinism plays a fundamental role. The main purpose<br />
of Task 5 is to study non-determinism in distributed computing. For example, non deterministic classes<br />
of complexity may be defined from the deterministic ones for distributed computing the same way the<br />
non-deterministic polynomial time class is defined from the polynomial deterministic time class. Hence,<br />
corresponding to the LD class (distributed languages that can be decided in constant time in the LOCAL<br />
model) we can consider the NLD, analogue non-deterministic class, for which there exists a certificate that<br />
can be verified in constant number of rounds in the LOCAL model. Note that as for the complexity classes<br />
and “yes-no” problems, several definitions are conceivable. Of course some questions about comparison<br />
between deterministic complexity classes and their non-deterministic versions arise naturally.<br />
At last, to complement this classical approach, this project envisions new computational paradigms for<br />
theory of distributed computing. We have identified two new paradigms relevant for distributed computing:<br />
distributed quantum computing and algorithmic game theory. Quantum-mechanical effects may have an<br />
interesting impact on the complexity and especially on its locality aspects for distributed computing [21].<br />
On the other hand, it is interesting to take into consideration game theory aspects in distributed computing.<br />
4 We prefer the term “yes-no” problem to “decision problem”, because “decision problem” has not the same meaning in<br />
wait-free or fault-tolerant context.<br />
7
Some classical problems, specifying the utilities of agents may be reformulated as equilibria in the sense of<br />
game theory.<br />
Clearly all these objectives are very challenging. For many of them, the current state of art is still rather<br />
primitive. Nevertheless, for most of them a lot of results already exist but generally are not stated in a general<br />
framework of complexity theory for distributed computing. Hence we think that most of the objectives are<br />
reachable and many new results and new insights into complexity theory for distributed computing are<br />
expected from this project.<br />
A crucial aspect of this proposal is that all objectives relate to both communities of Distributed Computing.<br />
Most of the time both communities adopt different approaches to tackle those issues. We expect<br />
the confrontation of these points of view to be valuable and to enable to converge to an outline of a general<br />
complexity and computability theory for distributed computing.<br />
3.2 Description by task<br />
3.2.1 Task 1: Project Management<br />
Coordinator: <strong>Carole</strong> Delporte-Gallet et <strong>Hugues</strong> Fauconnier (Paris)<br />
Local coordinators: David Ilcinkas (Bordeaux) et Achour Mostéfaoui (Rennes)<br />
The success of the project is crucial to an intensive collaboration between the three sites, involving<br />
frequent meetings and inter-site visits. DISPLEXITY is therefore planing four 2-day meetings every year<br />
during which all partners will present their most recent results related to the project. One of these meetings<br />
will be dedicated to the compilation of the results achieved during the year, to make sure that all partners<br />
synchronize. All dates for deliverables will fit with the dates of these annual special meetings.<br />
In addition, our budget requests a grant for a 1-week visit per year per permanent member (that can be<br />
used by the member himself or by his students). The coordinators of the project will pay a specific attention<br />
that frequent exchanges between the partners are done.<br />
The management of the web page of the project will play an important role. Indeed, the fundamental<br />
nature of the project requires that the results will be made available to the community as quickly as possible.<br />
Our web page will be a repository for preliminary versions of papers by the partners, enabling rapid diffusion<br />
of our preliminary results inside the project (via a private web site), and a rapid advertising of the finalized<br />
results outside the project (via a public web site).<br />
Scientific tasks<br />
Since the project is oriented toward fundamental research, the quality of the project will be evaluated by<br />
the ability of the partners to present and advertise their results in the most appropriate and prestigious<br />
conferences of the field.<br />
It should be noted that following the philosophy of the project, all the partners are included in all the<br />
tasks.<br />
3.2.2 Task 2: “yes-no” and Decision Problems in Distributed Computing<br />
Coordinator: Corentin Travers (Bordeaux)<br />
The overall objective of this task is the study of different settings for tackling decision problems in the<br />
various frameworks of distributed computing. As discussed in the general description of the project, we<br />
will mostly focus on a notion of decision defined as follows. Given a specification L (a specification can be<br />
formalized as a language-membership), deciding L is defined by the following two requirements:<br />
• If the inputs of the nodes satisfy L then all nodes must output “yes”;<br />
• Otherwise at least one node must output “no”.<br />
8
Hence, the “yes” answer must be unanimous. This definition is perfectly appropriated to, e.g., contexts<br />
in which any imperfection should result in a node raising an alarm [24]. On the other hand, one could<br />
strengthen the notion of decision by requiring all nodes to output “no” in case of an input not satisfying<br />
the given specification L. In other words, one could foresee settings in which both the “yes” and the “no”<br />
answers should be unanimous. While such a definition may not be appropriated in frameworks like LOCAL<br />
computation (because a 2-sided unanimous answer may yield communications between far away nodes) or<br />
wait-free computation (because consensus is not possible in this model [16]), there are frameworks for which<br />
such a definition is well suited. One typical example is mobile agent computing. Indeed, in this setting,<br />
it appears that any language that can be decided unanimously 1-sided can also be decided unanimously<br />
2-sided [20].<br />
In fact, one could define decision problems in many different manners. To give just a few examples,<br />
consider the following definitions:<br />
• Majority: If the inputs to the nodes satisfy L then a majority of nodes must output “yes”, otherwise<br />
a majority of nodes must output “no”.<br />
• Weighted: Every node i output a value x i ∈ [0, 1] so that if the inputs to the nodes satisfy L then<br />
∑<br />
i x i ≥ α, else ∑ i x i < α where α is some threshold that may depend on the size of the distributed<br />
system.<br />
• Functional: Every node i outputs a value x i belonging to some universe U so that if the inputs to the<br />
nodes satisfy L then f(x 1 , . . . , x n ) ≥ 0, else f(x 1 , . . . , x n ) < 0 where f : U ∗ → R.<br />
Objectives of the task:<br />
Sub-Task 2.1: Studying the pertinence of every decision model as a function of the computational model.<br />
The main computational models considered in this project are: the wait-free model, the LOCAL model,<br />
the CON GEST model, the population protocol model and the mobile agent model. One objective on<br />
Task 2 is, for each computational model, to determine the most appropriate decision model(s). The<br />
central question is: Is there a “natural” universal decision model?<br />
Sub-Task 2.2: Studying the respective powers of the decision models, for a fixed computational model.<br />
For instance, for any computational model, a language that can be decided in the 2-sided unanimous<br />
model can also be decided in the 1-sided unanimous-“yes” model. More generally, given two functions<br />
f and f ′ , and a fixed computational model, how to compare the ability of deciding according to f and<br />
the ability of deciding according to f ′ ?<br />
Sub-Task 2.3: Studying the relationship between the decision version of a problem and the computation<br />
version of the same problem. For instance, verifying the validity of a coloring can be achieved in<br />
one round in the LOCAL model, whereas computing a valid coloring requires a non-constant number<br />
of rounds. Preliminary studies indicate that computing is not always harder than verifying, at least<br />
in non-deterministic models. In particular, it is known that verification is sometimes harder than<br />
computation in the CON GEST model and in fault tolerant model. For example the verification of<br />
a consensus needs a stronger failure detector than the computation of the consensus. One objective<br />
on Task 2 is to tackle the relationship between computing and verifying under various computational<br />
models, and for various decision models.<br />
3.2.3 Task 3: Oracles<br />
Coordinator: Michel Raynal (Rennes)<br />
The main objective of this task is to evaluate the possibility and impossibility of what can be computed<br />
or more generally achieved in distributed systems. In a rather similar way to classical computation theory,<br />
the use of oracles is very useful to this end.<br />
9
Due to the impossibility result of FLP, oracles play a fundamental role concerning fault tolerance: as<br />
consensus is not possible in asynchronous models even with only one faulty process, in a very similar way<br />
to classical computability, it is natural to add oracles. Among other things, these oracles enable to establish<br />
hierarchies between (impossible) problems in distributed fault-tolerant computing. In this way, failure detectors<br />
are distributed oracles giving processes information about failures. The formal definition of failure<br />
detectors ensures that these oracles depend only on the failures and hence failure detectors encapsulate information<br />
about process crashes needed to solve the problem. One important point is that failure detectors<br />
may be compared by reduction: failure detector f is stronger than failure detector g if there is a distributed<br />
algorithm (a reduction) that enables to output g from f. In this way, failure detectors may be used to<br />
compare the difficulty of problems and enable to establish hierarchy between problems that are impossible<br />
to solve with process crashes.<br />
Another kind of oracles appears also in wait-free computations. Atomic objects like “test&set”, “cmp&swap”<br />
are not wait-free implementable in read-write restricted shared memory. In this way, calls to these objects<br />
may be considered as calls to oracles. Here, contrary to failure detectors that depend only on failures, the<br />
calls of these atomic objects depend also on the specification of the object. One of the main results about<br />
wait-free computation is the Herlihy’s consensus number [22] that enables to define a hierarchy between<br />
computational power of atomic objects.<br />
An important tool in wait-free computation is the BG simulation [4, 5] that enables to simulate systems<br />
t-resilient of n processes with systems of wait-free t + 1 processes. A very important question is how that<br />
kind of simulation may be extended to systems with oracles like failure detectors.<br />
More recently, another approach has been proposed([12]). Instead of having an oracle that gives information<br />
about failures, it is possible to consider that an adversary may choose which sets of processes are<br />
correct. For example, if we know that only processes a, b, c or a, b or a, c may be the processes that do not<br />
crash (the adversary may only choose one of these three sets) it is possible to have an algorithm solving<br />
consensus (intuitively a being correct in all the cases, a may realize the consensus). It is then possible to<br />
determine for a given problem some structural properties on the set of sets of correct processes that enable<br />
to solve this problem. For example, for the k-set agreement task the set of sets of correct processes for<br />
which there is a solution roughly corresponds to set of sets of corrects processes having at least k + 1 correct<br />
processes. A very interesting point is to study the relationship between failure detectors and the adversaries.<br />
In particular one of the main result concerning failure detectors, define the minimum failure detector to solve<br />
the k-set agreement, may be obtained by considering adversaries that enable to solve the k-set agreement.<br />
Concerning the more “network computing” oriented approach, oracles are also interesting. Some very<br />
simple oracles may also change the possibility and impossibility results. For example in an anonymous ring<br />
of processus, an oracle giving the number of nodes enables to separate rings of different sizes. More generally,<br />
the notion of oracles in this context may be an important tool concerning complexity issues.<br />
From a more practical point of view, many problems in distributed computing lead to graph representations<br />
where oracles may have an important role. For example, a peer to peer system is a graph, representing<br />
the topology of the logical network. Many relationships between entities can be represented by an edge in a<br />
graph: for instance a social network can be represented by a graph where there is an edge between 2 users<br />
if they are friends on Facebook, or an edge could be present between an item and a node if the node has<br />
used or tagged or bought an item (for example recommenders systems rely on such graphs). There are two<br />
main classes of problems that emerge: (i) either a graph exists and the goal is to extract some information<br />
from the graph that can then be used to provide a given functionnality in a distributed system (this includes<br />
studying the graph properteis of a social graph for example), or (ii) a graph is created to achieve some<br />
functionality of a system and its properties may help to tune the system (for example creation of a given<br />
overlay and study the properties of the resulting graph). This is typically some problems that should concern<br />
both communities.<br />
Objectives of the task:<br />
Sub-Task 3.1: Define a general theory of oracles in distributed computation is a preliminary goal of this<br />
task. For this the notion of reduction between oracles is central. The extension of usual reductions<br />
10
and simulations between systems to reductions and simulations of systems with oracles is another<br />
important goal of this task.<br />
Sub-Task 3.2: Concerning wait-free or fault-tolerant computation, the relationship between adversaries<br />
and oracles would enable to obtain new results. Some of these results may have practical impact and<br />
would help to refine the hierarchy between impossible problems.<br />
Sub-Task 3.3: An objective is to clearly define what are the graph properties expressed as oracles that<br />
are important in the context of distributed computing and to what extent they can help to improve<br />
the system. An another objective is to be able to capture such properties such the graph size, the<br />
community structure, the bottlenecks (for instance provided by centrality measure), etc.<br />
Sub-Task 3.4: Finally the previous objectives mentioned should be considered in the context of a dynamic<br />
and large-scale system, clearly leading to approximation. Calibrating such approximations and<br />
studying the impact on their utility is of crucial importance.<br />
3.2.4 Task 4: Complexity classes<br />
Coordinator: Pierre Fraigniaud (Paris)<br />
The overall objective of this task is to design significant complexity classes in the different frameworks of<br />
distributed computing. This design has to be made having in mind the general ”wish” of eventually being<br />
able to compare classes defined for different frameworks, e.g., to be able to compare a class defined for<br />
the wait-free model with a class defined for the LOCAL model. The definition of the complexity classes is<br />
subject to the notion of decidability one is considering. We foresee that the 1-sided unanimous “yes” notion<br />
is the most promising (cf. Task 2), and thus Task 4 is described according to this assumption. Nevertheless,<br />
it may be the case that Task 4 should be revisited if it occurs that our investigations performed in Task 2<br />
demonstrate that other notions of decidability should also be considered.<br />
The main basic complexity classes that will be investigated in the project are the following:<br />
• LD (for Local Decision) is the class of distributed languages that can be decided in constant time in<br />
the LOCAL model;<br />
• LDC is the class of distributed languages that can be decided in constant time in the CON GEST<br />
model;<br />
• WFD (for Wait-Free Decision) is the class of distributed languages that can be decided in the wait-free<br />
model;<br />
• MAD (for Mobile Agent Decision) is the class of distributed languages that can be decided in the<br />
mobile agent model;<br />
• POPD (for Population Protocol Decision) is the class of distributed languages that can be decided in<br />
the population protocol model.<br />
One note a significant difference between, on the one hand, the classes LD and LDC, and, on the other hand,<br />
the classes WFD, MAD and POPD. The former refers to a notion of time (number of rounds in the LOCAL<br />
or in CON GEST model). They have thus the flavor of computational complexity. The latter instead does<br />
not refer to any efficiency measure, and focuses solely on the ability of achieving decision. They have thus a<br />
flavor of decidability classes. Nevertheless, as it will be described in Task 5, there seems to be connections<br />
between these two very different types of classes, at least in the case of non-deterministic computation.<br />
It is also worth noticing that these complexity classes can be generalized to a probabilistic setting.<br />
For instance, one can define the class BPLD (for Bounded-error Probability Local Decision) as the set<br />
∪ 0≤p,q≤1 BPLD(p, q) where BPLD(p, q) is the class of all distributed languages that can be decided by a<br />
randomized distributed algorithm that runs in a constant number of communication rounds (in the LOCAL<br />
model) and produces correct answers on legal (respectively, illegal) instances with probability at least p<br />
(resp., q).<br />
11
Objectives of the task:<br />
Sub-Task 4.1: Characterizing the complexity classes listed above, or at least some fragments of them. For<br />
instance, it can be shown that by slightly reducing the definition, the languages in the class WFD<br />
are in one-to-one correspondence with covering spaces, in the sense of algebraic topology. In general,<br />
one objective of this sub-task is to determine whether these classes are decidable. Defining pertinent<br />
complexity measures for the wait-free model as well as for the mobile agent model is another objective.<br />
Notions like number of memory accesses, number of interactions or number of moves appear to be<br />
natural for the wait-free, population protocol and mobile agent model, respectively.<br />
Sub-Task 4.2: Studying the impact of randomization on the deterministic complexity classes LD, LDC,<br />
WFD, and MAD. For instance, it has been recently shown [19] that LD and BPLD coincide for a<br />
large range of values p and q, if one restricts the languages to the hereditary ones. The question<br />
”does randomization help?” is at the core of computer science. Proving or disproving LD = BPLD is<br />
probably beyond the objectives of this project. Nevertheless, we will work having this type of question<br />
in mind.<br />
Sub-Task 4.3: Defining appropriate notion of reductions that enable to compare the ”difficulty” of problems<br />
belonging to the same class. In the LOCAL model, a straightforward notion of reduction is to<br />
consider local computation mapping one instance to another while preserving the respective language<br />
membership. Defining the analog in, e.g., the framework of wait-free computation appears to be more<br />
challenging, because the notion of termination is more difficult to handle, and because faults can occur<br />
during the reduction.<br />
3.2.5 Task 5: Non-determinism in distributed computing<br />
Coordinator: Amos Korman (Paris)<br />
One can consider LD as the analogue of the class P, and then define the class NLD as the analogue of<br />
the class NP. Specifically, NLD (for non-deterministic local decision) is the class of distributed languages<br />
for which there exists a certificate (or proof ) that can be verified in a constant number of communication<br />
rounds in the LOCAL model. More precisely, NLD is the class of distributed languages L for which there<br />
exists a local algorithm such that:<br />
• if L ∈ L, then there exists a certificate c such that algorithm A applied on L with certificate c outputs<br />
“yes” for all nodes,<br />
• otherwise, for every certificate c, algorithm A applied on L with certificate c outputs “no” for at least<br />
one node.<br />
It is worth noticing that non-deterministic distributed computing does not find its interests in complexity<br />
theory only. In particular, the above notion of non-determinism is very much related to the area of computation<br />
with advice (e.g., [17, 18]), which aims at investigating the amount of information known to each<br />
node and its impact on the performances of the algorithm. In fact, the definition of NLD finds most of<br />
its similarities with the notion of proof labeling schemes (cf., e.g.,[23, 24]) which finds applications in the<br />
context of fault-tolerance and self-stabilization. Roughly, in certain contexts, one can afford spending lot of<br />
time computing the certificate, but once it is given to the nodes, one wants the verification to be fast. (One<br />
example is fault-detection in environments such as a plane or a nuclear plant). Clearly, LD ⊆ NLD, and it<br />
can be shown that LD ≠ NLD. Interestingly enough, it can be also shown that there exists a problem which<br />
is NLD-complete for the local reduction mentioned in Task 4. Still, the nature of NLD is far from being<br />
well-understood.<br />
Similarly, one can define non-deterministic versions of the classes LDC, WFD, and MAD. In the case of<br />
the mobile agent model, restricted to a single agent, it is form instance known that NMAD ∪ co-NMAD =<br />
MAD. As we will see, non-determinism finds all its interests when combined to the notion of oracles, which<br />
will be the objective of Task 3. Task 4 focuses on non-determinism per se.<br />
12
Objectives of the task:<br />
Sub-Task 5.1: One of our main concerns will be to identify the languages in NLD and NMAD. In particular,<br />
from a purely graph theoretic point of view, it is an intriguing question to identify which graph families<br />
belong to NLD. We are also aiming at identify subclasses of NLD defined according to the certificate<br />
size, finding connections between these subclasses, and finding complete problems for each of the<br />
subclasses.<br />
Sub-Task 5.2: Finding trade-offs between the certificate size and the locality of the verification. Indeed,<br />
the amount and type of information provided to each node is a central issue in the design of distributed<br />
algorithms and local algorithms in particular. It is however interesting to point out that most applications<br />
for proof labeling schemes do not perform in constant time. Thus, from the point of view of<br />
these applications, there is no real need to restrict the running time of the proof labeling schemes to<br />
constant. Reducing the certificate sizes of known proof labeling schemes at the price of somewhat<br />
increasing the locality of the verification is a challenging task that may thus contribute to the design<br />
of more compact applications. (For instance, doing this in the context of MST verification may yield<br />
breakthroughs in the strives for finding an efficient self-stabilizing MST construction protocol that is<br />
optimal in terms of memory).<br />
Sub-Task 5.3: Understanding the apparent lack of impact of non-determinism in the framework of wait-free<br />
computing. In other words, understanding why non-determinism does not seem to help in the context<br />
of wait-free computations. On the other hand, we have pointed out above that WFD can be essentially<br />
described in term of universal covers. It has been recently observed that output-free languages in<br />
NLD have strong connections with graph coverings. One objective of Task 5 will be to investigate the<br />
potential intriguing connections between WFD and NLD, which may not be a coincidence. Needless to<br />
say that establishing connections between classes defined for different distributed computational model<br />
would be a significant achievement.<br />
3.2.6 Task 6: New computational paradigms/frameworks<br />
Coordinator: Cyril Gavoille (Bordeaux)<br />
The principal objective of this task is to discover to what extent it is possible to improve the distributed<br />
complexity of combinatorial optimization problems by taking advantage of aspects of “new” computational<br />
paradigms. Inspired by recent developments of computational complexity, we propose to focus on two<br />
aspects: quantum-mechanical effects, and algorithmic game theory aspects.<br />
Our studies will involve different models of distributed computing, including processor network graphs,<br />
mobile-agent-based computing, and tasks involving the distribution of topological information (computing<br />
distance labelings, compact routing tables).<br />
The introduction of computational models based on quantum computing, starting from the works of<br />
Deutsch in the 1980’s [14], has led to the advent of a new branch of complexity theory. Independently of<br />
this, properties of quantum-mechanical systems have proven to be of interest from the perspective of game<br />
theory, information theory, and distributed systems [6, 13]. A major line of study concerns the application of<br />
quantum entanglement to reduce communication complexity, i.e., to decrease the number of communication<br />
bits required to solve a specific task performed within a system graph with several distributed agents [9].<br />
The influence of quantum information on the computing power of distributed systems with node anonymity<br />
and distributed systems in the presence of faults has also been studied for problems such as leader election<br />
or distributed consensus [28]. The goal of our project is to extend these results and obtain new insight which<br />
would be applicable in the context of distributed computing, especially in a combinatorial setting.<br />
In an orthogonal way, recent computational complexity have shown the necessity to revisit classical<br />
distributed algorithmic with the eyes of game theory. When conceiving a distributed system, it not sufficient<br />
to ask to all involved agents to act according to some algorithm, but also, as agents often correspond to<br />
various agents with their own economical aspects, to guarantee that it is indeed in their economical interest to<br />
13
do so. This often leads to add some game theoretic model to the system, specifying the utilities of involved<br />
agents, and then revisiting whether situations computed globally correspond to equilibria in the sense of<br />
game theory.<br />
Objectives of the task:<br />
Sub-Task 6.1: Reducing the time complexity of distributed algorithms using quantum effects. The starting<br />
point for our considerations is the well-established LOCAL model of distributed computing with a<br />
network graph of processors exchanging messages of unbounded size. In this context, we wish to discover<br />
how the introduction of a globally pre-entangled state of the system or the application of quantum<br />
communication channels can decrease the number of rounds required to solve classical combinatorial<br />
optimization problems, such as graph (∆ + 1)-coloring. Currently, only some very simple proof-ofconcept<br />
examples of such problems are known [21]. We are especially interested in effects which<br />
require quantum entanglement and cannot be obtained using purely classical correlations, such as<br />
shared random bits.<br />
Sub-Task 6.2: Providing a comparison of the “computational power” of the quantum and non-quantum<br />
models, formalising the notion of locality in quantum distributed computing, and showing how it<br />
essentially differs from the understanding of locality in the LOCAL model. In particular, we would<br />
like to obtain more precise algorithmic characterizations of probability distributions of feasible outputs<br />
which can be obtained by a quantum system in a given number of rounds. We also intend to elaborate<br />
lower bounds on the complexity of problems in quantum models by making use of more powerful models<br />
of computations based on stronger-than-quantum non-local boxes (Popescu-Rohrlich boxes and their<br />
variants), which we intend to define for the purpose.<br />
Sub-Task 6.3: Identifying the potential of quantum information in topological queries. We would like to<br />
study the potential of quantum information in reducing the size of distributed data structures required<br />
for tasks such as adjacency labeling or distance labeling of a graph, allowing the recovery of information<br />
about the location of a pair (or more generally subset) of nodes of a graph based only on the information<br />
stored within these nodes. In this sense, we would like to establish if information “spread out” over the<br />
whole graph using entangled states can prove more efficient in local queries than classical information<br />
(without globally violating the information bound implied by Holevo’s theorem).<br />
Sub-Task 6.3: Providing a comparison of the “computational power” of models when rationality is imposed<br />
and when it is not, formalising when local rules in distributed computing correspond to rational rules<br />
in the sense of game theory, and showing how it may or may not influence the computational power<br />
of the involved models. In particular, we would like to understand how complexity classes proposed<br />
in the other tasks differ or not from their analog when the hypothesis of rationaly is imposed to<br />
agents. This involves to consider the various possibilities to go from a game to a distributed system<br />
(ex. by repeating local games, by evolutionay game theory models) and then characterize what can be<br />
computed by which model using the corresponding notion of rationality.<br />
3.3 Tasks schedule, deliverables and milestones<br />
All the tasks listed in the previous sections are in symbiosis, and there is no precedence constraints among<br />
them. Each of the tasks is related to a well identified problem, and the resolution of any problem will benefit<br />
to many others. In fact, the tasks are as many different ways of tackling key fundamental problems in the<br />
framework of distributed computing. Moreover, because of the theoretical research flavor of the tasks, it is<br />
quite hard (and probably inappropriate) to decide when a task should start and end.<br />
On the other hand, DISPLEXITY will pay attention to allow the participants to the project to be<br />
cross fertilized by the results of all its members. This will be implemented via frequent meetings and visits<br />
between the three sites. By doing so, the participants to DISPLEXITY will share a common expertise, and<br />
be able to solve more and more complex problems related to the tasks of the project. Within the 4-years<br />
14
Figure 1: Interaction graph beween partners (DBLP).<br />
duration of the project, it is therefore expected that many key problems will have been solved, providing as<br />
many breakthroughs in the field of distributed computing.<br />
One of these meetings will be dedicated to the compilation of the results achieved during the year. The<br />
dates for deliverables (t+12, t+24, t+36, t+48) containing the progress of each task will fit with the dates<br />
of these annual special meetings.<br />
4 Dissemination and exploitation of results, intellectuel property<br />
The main expected results of this project are scientific advances in the investigated research fields. Research<br />
results will be presented at the leading international conferences and in the leading scientific journals.<br />
Each year, PODC, the top level ACM conference on distributed computing invites proposals for workshops<br />
that are closely related to the scope of the conferences, i.e. on topics related to the theory, design, specification<br />
or implementation of distributed systems. Workshops are in conjunction with the conference on the day<br />
immediately preceding or following the conference program. The organization of such a workshop will be<br />
an opportunity to promote our scientific results and to draw the attention of the international community.<br />
And we plan to propose such a workshop the third year of his project.<br />
DISPLEXITY will maintain a project Web Site that will make available the project results; it will be<br />
a useful portal open to all researchers for accessing papers and deliverables.<br />
15
5 Consortium description<br />
5.1 Partners description and relevance, complementarity<br />
The consortium gathers members from three French leading labs in Distributed Computing (LIAFA, LaBRI,<br />
and IRISA). Each of these labs has a worldwide reputation of excellence in the design and analysis of<br />
distributed algorithms. DISPLEXITY has paid attention to include experts of most sub-topics in distributed<br />
computing, including network computing, mobile computing, fault-tolerant computing, population protocols,<br />
failure detector theory, etc. In order to illustrate the high level of expertise of the DISPLEXITY partners,<br />
let us just mention that the consortium includes several ACM PODC and DISC Program Committee Chairs<br />
(which are the top Int’l conferences in the domain), several DISC and SIROCCO Steering Committee Chairs,<br />
and regular Program Committee members of most conferences in distributed computing (ACM PODC, DISC,<br />
SIROCCO, OPODIS, SSS, IEEE ICDCS, IPDPS, etc.) and Editorial Board members of top Int’l journals<br />
(e.g., IEEE Transactions on Computers, IEEE Transactions on Parallel and Distributed Systems, Journal of<br />
Parallel and Distributed Computing).<br />
The quality of the consortium is more than the sum of the quality of its members. Indeed, the main<br />
interest of DISPLEXITY will be its ability to tackle the main research models in distributed computing in<br />
parallel. To enable DISPLEXITY to take up the challenges of dealing with apparently so many different computational<br />
models, we have focused our attention on yes/no problems, which will ensure a global coherency<br />
of the project. This apparent narrowing of the problem field is actually not restrictive as it is sufficient for<br />
the development of a complexity theory (this is because such theory is based on decision problems).<br />
As illustrated by the collaborative diagram displayed on Figure 1, the consortium includes two well<br />
identified sub-groups, with each of the sub-groups scattered over the three collaborative institutions LIAFA-<br />
LaBRI and IRISA. Roughly, one sub-group includes individuals dealing with problems related to asynchrony<br />
and faults, while the other one includes partners dealing with network computing.<br />
It should be mentioned that many participants regularly attend to the same conferences gathering both<br />
communities. The high added value brought by DISPLEXITY is to provide a general framework (Computational<br />
Complexity) in which these people will exchange and collaborate. Conversely, the richness of the<br />
consortium, beyond the individual quality of its members, is to gather these two groups together: each one<br />
separately will not be able to cover more than a fragment of the theory. We do hope, and actually strongly<br />
believe, that the same drawing at the end of DISPLEXITY will display a single component, with connecting<br />
edges corresponding to many significative publications in Distributed Computational Complexity.<br />
Bordeaux (LABRI): The LaBRI (Laboratoire Bordelais de Recherche en Informatique) is a research<br />
unit associated with the CNRS (UMR 5800), the University of Bordeaux 1, the IPB and the University<br />
of Bordeaux 2. It has significantly increased in staff numbers over recent years and now includes around<br />
350 members (academics, researchers, PhDs, etc.). Today the members of the laboratory are grouped in<br />
six teams, each one combining basic research, applied research and technology transfer : * Combinatorics<br />
and Algorithmics * Image and Sound * Languages, Systems and Networks * Formal Methods * Models and<br />
Algorithms for Bio-informatics and Data Visualisation * Supports and Algorithms for High Performance<br />
Numerical Applications<br />
The LaBRI participants to DISPLEXITY are all members of the Combinatorics and Algorithmics team<br />
and most of them are in the INRIA team CEPAGE. They are all recognized worldwide experts in distributed<br />
computing. The LaBRI team includes specialists on various models and topics in the scope of the project,<br />
in particular fault tolerance issues, network and communication algorithmics, and (possibly quantum) distributed<br />
information.<br />
Paris (LIAFA): The LIAFA (Laboratoire d’Informatique Algorithmique: Fondements et Applications)<br />
is supported jointly by the French National Center for Scientific Research (CNRS) and by the University<br />
Paris Diderot - Paris 7. It is member of the Fondation Sciences mathématiques de Paris. The main research<br />
topics addressed by LIAFA are related to theoretical computer science. The one hundred members<br />
16
of LIAFA (academics, researchers, and PhDs) are divided in five research teams: Algorithms and Complexity,<br />
Combinatorics, Distributed Algorithms and Graphs, Automata and applications, and Modeling and<br />
verification.<br />
DISPLEXITY involves all members of the team Distributed Algorithms and Graphs whose research<br />
interests are dealing with distributed computing. All these people are worldwide experts in distributed<br />
computing, covering most of the topics in the field, both theoretically-oriented and practically-oriented<br />
(they are all members of the INRIA team-projet ”GANG”).<br />
Rennes (IRISA-ASAP): The ASAP (As Scalable As Possible) team, lead by Anne-Marie Kermarrec,<br />
is part of IRISA/INRIA Rennes a research unit associated mainly with University of Rennes, CNRS and<br />
INRIA. The research activities of ASAP range from theoretical foundations to practical protocols and implementations<br />
for (mainly large-scale and dynamic) distributed systems in order to cope with the recent<br />
and tremendous evolution of distributed systems. Effectively we observed huge evolutions: (i) Scale shift in<br />
terms of system size, geographical spread, volume of data, and (ii) Dynamic behaviour due to versatility,<br />
mobility, connectivity patterns.<br />
The research of the ASAP Project-Team is along two main themes: Distributed computing models and<br />
abstractions and Peer-to-peer distributed systems and applications. These research activities encompass<br />
both long term fundamental research seeking significant conceptual advances, and more applied research<br />
to validate the proposed concepts against real applications. ASAP regroups 14 researchers (including one<br />
professor, one senior researcher, two associate professors and one researcher).<br />
The proposal also involves one researcher from team Algorithmic and Complexity from Computer Science<br />
Laboratory (LIX) from Ecole Polytechnique. The research activity of this member is engaged in research<br />
along two themes related to this proposal: models of computation, and in particular distributed models of<br />
computation and algorithmic game theory.<br />
5.2 Qualification of the proposal coordinator<br />
<strong>Carole</strong> Delporte-Gallet and <strong>Hugues</strong> Fauconnier will share the coordination task. They have a long-standing<br />
collaboration in research projects and have a strong experience as investigators of projects and organizators<br />
of workshops on Distributed Computing. They are top level experts in Fault Tolerance Distributed Computing<br />
and they published papers in many journals like JACM, Distributed Computing, TOPLAS and in top<br />
level conferences of this aera (PODC, DISC, DSN, ICDCS, ...). They have been program committee members<br />
of established conferences in Distributed Computing such as PODC, DISC, IEEEE ICDCS, OPODIS,<br />
SIROCCO...<br />
As investigators of projects, they have been investigators in the last past years of the Action Spécifique:<br />
Algorithmes Distribués (03-04) and of the BQR-action “Distributed Quantum computing”of the University<br />
Paris-Diderot University’ (2010). They organized a school on Distributed Computing in 2003 and two<br />
workshops: “Algorithmique distribuée et applications” in 2004 and ”Distributed Quantum computing” in<br />
2010.<br />
C. Delporte-Gallet is Professor in computer science at the University Paris-Diderot and leader of the<br />
team “Distributed Algorithms andGraphs” of the LIAFA . She was alumni of the ENS Sèvres (79-83). She<br />
holds her Ph.D. in Computer Science from the University Paris Diderot in 1983 and is HDR in 2001.<br />
H. Fauconnier received his Ph.D. in 1982 and HDR degree in 2001 in Computer Science from the University<br />
Paris-Diderot, after Master degrees in Mathematics and Computer Science. He is maitre de conferences<br />
in Computer Science at the University Paris-Diderot.<br />
<strong>Hugues</strong> Fauconnier was local investigator for the ACI project FRAGILE (2006-2008) and for the ANR-<br />
VERSO SHAMAN (2008-2012). It is also local investigator of an Ile de France Post-Doc program PEFI-<br />
CAMO (2009-2011).<br />
17
5.3 Qualification and contribution of each partner<br />
Partner 1: Paris(Coordinator)<br />
Name<br />
First<br />
name<br />
Position<br />
Delporte <strong>Carole</strong> PR Distributed computing, fault<br />
tolerance (t-resilent, wait free,<br />
failure detector, adversaries),<br />
population protocol<br />
Fauconnier <strong>Hugues</strong> MCF Distributed computing, fault<br />
tolerance (t-resilent, wait free,<br />
failure detector, adversaries),<br />
population protocol<br />
Fraigniaud Pierre DR Distributed computing, Network<br />
computing, Mobile computing<br />
Field of research PM Contribution to the proposal (4<br />
lignes max)<br />
36 Coordinator of the<br />
project(Task1)<br />
36 Coordinator of the project<br />
(Task1)<br />
36 Coordinator Task 4<br />
Korman Amos CR Distributed computing, labeling<br />
36 Coordinator Task 5<br />
schemes, sparse spannners<br />
Viennot Laurent DR Networks 12<br />
Bauman Hervé PhD Rumor spreading 12<br />
Koegler Xavier PhD Population protocol 24<br />
Arfaoui Heger PhD Distributed computing complexity<br />
36<br />
in the LOCAL<br />
model<br />
18
Partner 2: Rennes<br />
Name<br />
First<br />
name<br />
Position<br />
Mostefaoui Achour MCF Distributed computing. Implementation<br />
of Oracles and design<br />
of data structures well suited to<br />
a given oracle<br />
Raynal Michel PR Distributed computing. Decidabiliy<br />
of decision tasks and<br />
lower bounds for implementation<br />
in a given system<br />
Kermarrec<br />
Anne-<br />
Marie<br />
Field of research PM Contribution to the proposal (4<br />
lignes max)<br />
DR Peer-to-peer Systems. Definition<br />
and implementation of oracles<br />
well-suited to peer-to-peer<br />
systems<br />
Bournez Olivier PR Computability and complexity<br />
with distributed models, anonymous<br />
models, analog models.<br />
Algorithmic game theory.<br />
Imbs Damien Phd<br />
Student<br />
Distributed computing. Decidabiliy<br />
of decision tasks and<br />
lower bounds for implementation<br />
in a given system<br />
36 local coordinator (Task 1)<br />
36 coordinator Task 3<br />
12<br />
24<br />
24<br />
19
Partner 3: LABRI<br />
Name<br />
First<br />
name<br />
Position<br />
Field of research PM Contribution to the proposal (4<br />
lignes max)<br />
Gavoille Cyril PR Distributed Graph Algorithms, 36 Task 6 Coordinator<br />
Quantum Information<br />
Ilcinkas David CR Distributed Graph Algorithms, 36 Local coordinator (Task 1)<br />
Mobile Agent Computing<br />
Johnen Colette PR Fault Tolerance 28<br />
Klasing Ralf DR Network algorithms, Mobile 24<br />
Agent Computing<br />
Kosowski Adrian CR Distributed Comptuting, Quantum<br />
20<br />
Information<br />
Travers Corentin MCF Fault Tolerance 36 Task 2 Coordinator<br />
Halftermeyer Pierre PhD Disributed Distance Oracles 12<br />
Wade Ahmed PhD Mobile Agent Computing, Dynamic<br />
Graphs<br />
24<br />
20
6 Scientific justification of requested ressources<br />
The budget concerning small material will mostly be used for providing participants with appropriate computers,<br />
and to give adequate resources to the PhDs and Post-doc involved in the project.<br />
A large part of the requested grant concerns missions. For the success of the project, it is indeed crucial<br />
that the participants get enough supports for frequent exchanges and visits between partners. In addition,<br />
the topics addressed by the project are quite hot, and evolve rapidly. It is thus very important that the<br />
participants of DISPLEXITY can participate to the most relevant conferences of the field, for presenting<br />
their results, and to carry on collaborations with colleagues. The yearly budget for missions is actually<br />
increasing with time, it is planed that the need for traveling will increase with time, for participants are<br />
expected to present their new results in various places, including the most visible scientific events.<br />
Most of the partners of this project plan to concentrate their research on DISPLEXITY during the four<br />
next years.<br />
6.1 Partner 1: Paris<br />
Personnel<br />
PhD<br />
Master Internship<br />
Amount<br />
3 years<br />
4* 6 months<br />
Mission 178 000<br />
Small equipment 20 000<br />
• Equipment<br />
None.<br />
• Staff<br />
PhD LIAFA: “yes-no decision problems”<br />
This PhD position is related to Task 2 (“yes-no” and Decision Problems in Distributed Computing).<br />
Its first objective is, for the wait-free model and the LOCAL computational model, to determine<br />
the most appropriate decision model. The second objective will be to study the respective powers<br />
of the decision model(s), for a fixed computational model and the relationship between the decision<br />
version of a problem and the computation version of the same problem. For instance, verifying the<br />
validity of a coloring can be achieved in one round in the LOCAL model, whereas computing a valid<br />
coloring requires a non-constant number of rounds. For example the verification of a consensus needs<br />
a stronger failure detector than the computation of the consensus. One objective of Task 2 is to tackle<br />
the relationship between computing and verifying under various computational models, and for various<br />
decision models.<br />
This PhD position will be open to the best Master students in France and the aforementioned foreign<br />
universities.<br />
Master student internship Master student internship will be proposed in relation with the different<br />
aspects of the project and more precis task 3, 4, 5 and 6. It is expected to supervise one master student<br />
per year for a six months period each (the typical internship length in our university).<br />
• Subcontracting None<br />
• Mission<br />
The total amount for travel is 178 000 and is justified as follows:<br />
– As pointed out in the Task 1, four 2-days workshops will be organized each year which induce a<br />
cost of 4 500 Euros per such meeting for all members. The resulting cost is 76 000.<br />
21
– As pointed out in the Task 1, our budget requests a grant for a 1-week visit per year per permanent<br />
member for a total cost of 20 000 Euros.<br />
– It is expected that most of the results will be communicated to top level international conferences<br />
in distributed computing. Moreover, as explained at the beginning, it is also vitally important that<br />
the members collaborate on the project topic with the other (international) experts on the domain.<br />
The resulting cost is roughly estimated to 82000 Euros. It could be considered as rather high,<br />
but it should be taken into account that the permanent members of the Paris partner published<br />
82 papers in the major conferences of the domain in the last four years (the same duration as<br />
the project). Even without taking into consideration the collaborations with other colleagues,<br />
the participation to major workshops without refereed publications, and the missions of the non<br />
permanent staff, this could already almost justify the value of 82 000 Euros for missions.<br />
• Costs justified by internal invoicies<br />
• Other expenses<br />
The budget concerning small material will mostly be used for providing participants with appropriate<br />
computers, and to give adequate resources to the PhDs and Post-doc involved in the project. Various<br />
expenditures will be made throughout the project duration for a total cost of 20 000 Euros.<br />
6.2 Partner 2: Rennes<br />
Personnel<br />
Post-doc<br />
Master Internship<br />
Mission<br />
Small equipment<br />
Amount<br />
18 months<br />
4*6 months<br />
112 000 E<br />
10 000 E<br />
• Equipment<br />
None.<br />
• Staff<br />
The post-doc student will be hired for working mainly on Task 3 devoted to Oracles. Of course, she will<br />
also participate to the other tasks. The notion of oracle has been introduced many years ago mainly for<br />
solving the consensus problem is asynchronous systems but has been extended to tackle many other<br />
decision problems and among others we propose to extend this notion to the graph topology that<br />
underlies any computing network.<br />
The master student internships will be proposed to students in order to work on specific aspects on<br />
the design of distributed data structures that rely on some oracle. These aspects will be connected to<br />
tasks 5, 6 and 3. The internships will be proposed on a one or two per year basis probably starting<br />
from the second year of the project.<br />
• Subcontracting<br />
None<br />
• Mission<br />
The total amount for travel is 112000 euros and is justified as follows:<br />
– As pointed out in the Task 1, our budget requests a grant for a 1-week visit per year per member<br />
for a total cost of 20000 Euros.<br />
22
– As pointed out in the Task 1, four 2-days workshops will be organized each year which will induce<br />
a cost of 1 250 Euros per such meeting for all members. The resulting cost is 20 000.<br />
– It is expected that most of the results will be communicated to top level international conferences<br />
in distributed computing. Moreover, as explained at the beginning, it is also vitally important that<br />
the members collaborate on the project topic with the other (international) experts on the domain.<br />
The resulting cost is roughly estimated to 30 000 Euros. It could be considered as rather high,<br />
but it should be taken into account that the permanent members of the Rennes partner published<br />
more than a hundred papers in the major conferences of the domain in the last four years (the<br />
same duration as the project). Even without taking into consideration the collaborations with<br />
other colleagues, the participation to major workshops without refereed publications, and the<br />
missions of the non permanent staff, this could already almost justify the value of 72 000 Euros<br />
for missions (on a basis of two int. confrences per permanent member per year).<br />
• Costs justified by internal invoicies<br />
• Other expenses<br />
The amount of 10000 Euros for small equipment is for purchasing four laptops. One for the post-doc<br />
student, one for the different master students over four years and the two others are for a renewal of<br />
machines for two of the members of the team involved in this ANR project.<br />
6.3 Partner 3: Bordeaux<br />
Personnel<br />
Post-doc<br />
Master Internships<br />
Mission<br />
Small equipment<br />
Amount<br />
1.5 years<br />
20 months<br />
164000 Euros<br />
15000 Euros<br />
• Equipment<br />
None.<br />
• Staff<br />
Post-doc position. The 18-month post doc will support us in Task 4 dedicated to the complexity<br />
classes and will particularly focus on studying the impact of randomization on the deterministic complexity<br />
classes (Task 4.2). A strong expertise in complexity theory and randomized algorithms will<br />
be requested in the position description. The latter will be disseminated worldwide through the main<br />
e-mailing list of the theoretical computer science community. This post-doc induces a cost of 1.5*42000<br />
= 63000 Euros.<br />
Master student internships. Master student internships will be proposed in relation with the<br />
different aspects of the project. It is expected to supervise one master student per year for a fivemonth<br />
period each (the typical internship length in our university). This induces a cost of 4*5*417,09<br />
= 8341.80 Euros.<br />
• Subcontracting None.<br />
• Mission<br />
The total amount for travel is 164000 Euros and is justified as follows:<br />
– As pointed out in Task 1, four 2-days workshops will be organized each year which induce a cost<br />
of 3500 Euros per such meeting for all members. The resulting cost is 56000 Euros.<br />
23
– As pointed out in Task 1, our budget requests a grant for a 1-week visit per year per permanent<br />
member for a total cost of 24000 Euros.<br />
– It is expected that most of the results will be communicated to top level international conferences<br />
in distributed computing. Moreover, as explained at the beginning, it is also vitally important<br />
that the members collaborate on the project topic with the other (international) experts on the<br />
domain. The resulting cost is roughly estimated to 84000 Euros. It could be considered as rather<br />
high, but it should be taken into account that the permanent members of the Bordeaux partner<br />
published 84 papers in the major conferences of the domain in the last four years (the same<br />
duration as the project). Even without taking into consideration the collaborations with other<br />
colleagues, the participation to major workshops without refereed publications, and the missions<br />
of the non permanent staff, this could already almost justify the value of 84000 Euros for missions.<br />
• Costs justified by internal invoicies<br />
None.<br />
• Other expenses<br />
The budget concerning small material will mostly be used for providing participants with appropriate<br />
computers, and to give adequate resources to the PhDs and Post-doc involved in the project. Various<br />
expenditures will be made throughout the project duration for a total cost of 15000 Euros.<br />
7 Annexes<br />
7.1 References<br />
References<br />
[1] Dana Angluin, James Aspnes, Zoë Diamadi, Michael J. Fischer, and René Peralta. Computation in<br />
networks of passively mobile finite-state sensors. Distributed Computing, 18(4):235–253, 2006.<br />
[2] Franois Bonnet and Michel Raynal. Anonymous asynchronous systems: The case of failure detectors.<br />
In DISC 2010, pages 206–220, 2010.<br />
[3] Franois Bonnet and Michel Raynal. Consensus in anonymous distributed systems: Is there a weakest<br />
failure detector? In AINA 2010, pages 206–213, 2010.<br />
[4] Elizabeth Borowsky and Eli Gafni. A simple algorithmically reasoned characterization of wait-free<br />
computations (extended abstract). In PODC, pages 189–198, 1997.<br />
[5] Elizabeth Borowsky, Eli Gafni, Nancy A. Lynch, and Sergio Rajsbaum. The bg distributed simulation<br />
algorithm. Distributed Computing, 14(3):127–146, 2001.<br />
[6] Anne Broadbent and Alain Tapp. Can quantum mechanics help distributed computing? ACM SIGACT<br />
News - Distributed Computing Column, 39(3):67–76, September 2008.<br />
[7] Tushar Deepak Chandra, Vassos Hadzilacos, and Sam Toueg. The weakest failure detector for solving<br />
consensus. J. ACM, 43(4):685–722, July 1996.<br />
[8] Tushar Deepak Chandra and Sam Toueg. Unreliable failure detectors for reliable distributed systems.<br />
J. ACM, 43(2):225–267, March 1996.<br />
[9] Richard Cleve and Harry Buhrman. Substituting quantum entanglement for communication. Physical<br />
Review A, 56(2):1201–1204, Aug 1997.<br />
24
[10] <strong>Carole</strong> Delporte-Gallet, <strong>Hugues</strong> Fauconnier, and Rachid Guerraoui. Tight failure detection bounds on<br />
atomic object implementations. J. ACM, 57(4), 2010.<br />
[11] <strong>Carole</strong> Delporte-Gallet, <strong>Hugues</strong> Fauconnier, Rachid Guerraoui, and Anne-Marie Kermarrec. Brief announcement:<br />
byzantine agreement with homonyms. In Friedhelm Meyer auf der Heide and Cynthia A.<br />
Phillips, editors, SPAA, pages 74–75. ACM, 2010.<br />
[12] <strong>Carole</strong> Delporte-Gallet, <strong>Hugues</strong> Fauconnier, Rachid Guerraoui, and Andreas Tielmann. The disagreement<br />
power of an adversary. Distributed Computing, 2010. Accepted in DISC: Special issue.<br />
[13] Vasil S. Denchev and Gopal Pandurangan. Distributed quantum computing: A new frontier in distributed<br />
systems or science fiction? ACM SIGACT News - Distributed Computing Column, 39(3):77–95,<br />
September 2008.<br />
[14] David Deutsch. Quantum theory, the Church-Turing principle and the universal quantum computer.<br />
Proceedings of the Royal Society of London, A400:97–117, 1985.<br />
[15] Shlomi Dolev. Self-Stabilization. MIT Press, 2000.<br />
[16] Michael J. Fischer, Nancy A. Lynch, and Michael S. Paterson. Impossibility of distributed consensus<br />
with one faulty process. J. ACM, 32(2):374–382, April 1985.<br />
[17] Pierre Fraigniaud, Cyril Gavoille, David Ilcinkas, and Andrzej Pelc. Distributed computing with advice:<br />
Information sensitivity of graph coloring. In 34 th International Colloquium on Automata, Languages and<br />
Programming (ICALP), volume 4596 of Lecture Notes in Computer Science, pages 231–242. Springer,<br />
July 2007.<br />
[18] Pierre Fraigniaud, Amos Korman, and Emmanuelle Lebhar. Local mst computation with short advice.<br />
In Phillip B. Gibbons and Christian Scheideler, editors, SPAA, pages 154–160. ACM, 2007.<br />
[19] Pierre Fraigniaud, Amos Korman, and David Peleg. Local distributed decision. CoRR, abs/1011.2152,<br />
2010.<br />
[20] Pierre Fraigniaud and Andrzej Pelc. Decidability classes for mobile agents computing. CoRR,<br />
abs/1011.2719, 2010.<br />
[21] Cyril Gavoille, Adrian Kosowski, and Marcin Markiewicz. What can be observed locally? round-based<br />
models for quantum distributed computing. Technical report, arXiv: quant-ph/0903.1133, 2009.<br />
[22] Maurice P. Herlihy. Wait-free synchronization. ACM Transactions on Programming Languages and<br />
Systems, 13(1):123–149, January 1991.<br />
[23] Amos Korman and Shay Kutten. Distributed verification of minimum spanning trees. Distributed<br />
Computing, 20(4):253–266, 2007.<br />
[24] Amos Korman, Shay Kutten, and David Peleg. Proof labeling schemes. Distributed Computing,<br />
22(4):215–233, 2010.<br />
[25] Nathan Linial. Locality in distributed graphs algorithms. SIAM Journal on Computing, 21(1):193–201,<br />
1992.<br />
[26] Moni Naor and Larry J. Stockmeyer. What can be computed locally? SIAM J. Comput., 24(6):1259–<br />
1277, 1995.<br />
[27] David Peleg. Distributed Computing: A Locality-Sensitive Approach. SIAM Monographs on Discrete<br />
Mathematics and Applications, 2000.<br />
[28] Seiichiro Tani, Hirotada Kobayashi, and Keiji Matsumoto. Exact quantum algorithms for the leader<br />
election problem. In 22 nd Annual Symposium on Theoretical Aspects of Computer Science (STACS),<br />
volume 3404 of Lecture Notes in Computer Science, pages 581–592. Springer, February 2005.<br />
25
7.2 CV, resume<br />
26
Partner 1: Paris<br />
27
<strong>DELPORTE</strong>-<strong>GALLET</strong> <strong>Carole</strong><br />
Coordinator<br />
Age: 51<br />
Position: PR<br />
Email: delporte@liafa.jussieu.fr<br />
URL: www.lliafa.jussieu.fr/˜cd<br />
LIAFA, Université Paris Diderot<br />
Case 7014<br />
F-75205 Paris cedex 13<br />
Tel: (+33) 1 57 27 92 25<br />
Cursus<br />
PhD (1983) and HDR (2001) from University Denis Diderot<br />
Publications<br />
Number of publications in refereed international journals: 10<br />
(JACM, TOPLAS, Distributed computing, Information and Computation, Information and Control, JPDC...<br />
Number of publications in refereed international conferences with proceedings: 39<br />
(PODC, DISC, DSN, ICDCS, ICALP, SPAA, SRDS, OPODIS, SIROCCO, IPDPS, ICDCN, DCOSS....)<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• C.Delporte-Gallet, H.Fauconnier, R.Guerraoui and A. Tielmann. The Weakest Failure Detector for<br />
Message Passing Set-Agreement. Distributed computing, online . (2010)<br />
• M.K. Aguilera,C.Delporte-Gallet, H.Fauconnier and S. Toueg. Partial synchrony based on set timeliness.<br />
PODC, pages 102-110. (2009)<br />
• C.Delporte-Gallet, S. Devismes and H.Fauconnier Stabilizing leader election in partial synchronous<br />
systems with crash failures. J. Parallel Distrib. Comput., volume 70 (1), pages 45-58,. (2010)<br />
• C.Delporte-Gallet, H.Fauconnier, R.Guerraoui and A. Tielmann. The Weakest Failure Detector for<br />
Message Passing Set-Agreement. DISC, pages 109-120 ). (2009)<br />
• C.Delporte-Gallet, H.Fauconnier, F. Freiling, L. Penso and A. Tielmann. From Crash-Stop to Permanent<br />
Omission: Automatic Transformation and Weakest Failure Detectors. DISC, pages<br />
165-178. (2008)<br />
Prices, distinctions: Best paper award DISC 2009.<br />
28
FAUCONNIER <strong>Hugues</strong><br />
Coordinator<br />
Age:55<br />
Position: MC<br />
Email: fauconnier@liafa.jussieu.fr<br />
URL: www.lliafa.jussieu.fr/˜hf<br />
LIAFA, Université Paris Diderot<br />
Case 7014<br />
F-75205 Paris cedex 13<br />
Tel: (+33) 1 57 27 92 25<br />
Cursus<br />
PhD (1982) and HDR (2001) from University Denis Diderot<br />
Publications<br />
Number of publications in refereed international journals: 10<br />
(JACM, TOPLAS, Distributed computing, TCS, Information and Computation, JPDC,TPDS.....)<br />
Number of publications in refereed international conferences with proceedings: 38<br />
(PODC, DISC, DSN, ICDCS, SRDS, SPAA, OPODIS, SIROCCO, IPDPS, ICDCN,DCOSS....)<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• C.Delporte-Gallet, H.Fauconnier and R. Guerraoui. Tight failure detection bounds on atomic object<br />
implementations. J. ACM, volume 57 (4). (2010)<br />
• M.K. Aguilera,C.Delporte-Gallet, H.Fauconnier and S. Toueg. On implementing Omega in systems<br />
with weak reliability and synchrony assumptions. Distributed Computing, volume 21 (4), pages<br />
285-314. (2010)<br />
• C.Delporte-Gallet, H.Fauconnier, R. Guerraoui and A. Tielmann. Fault-Tolerant Consensus in Unknown<br />
and Anonymous Networks. ICDCS, pages 368-375). (2009)<br />
• C.Delporte-Gallet, H.Fauconnier, R. Guerraoui and A. Tielmann. The Disagreement Power of an<br />
Adversary. DISC(Best Paper), pages 8-21). (2009)<br />
• C.Delporte-Gallet, H.Fauconnier and R. Guerraoui. Sharing is harder than agreeing. PODC, pages<br />
85-94. (2008)<br />
Prices, distinctions: Best paper award DISC 2009.<br />
29
FRAIGNIAUD Pierre<br />
Age: 48<br />
Position: Directeur de Recherches CNRS<br />
Email: pierre.fraigniaud@liafa.jussieu.fr<br />
URL: www.liafa.jussieu.fr/˜pierref<br />
LIAFA, Université Paris Diderot<br />
Case 7014<br />
F-75205 Paris cedex 13<br />
Tel: (+33) 1 57 27 94 00<br />
Cursus<br />
PhD (1990) and Habilitation (1994) from ENS Lyon<br />
Other professional experiences:<br />
P. Fraigniaud is Director of LIAFA (Laboratoire d’Informatique Algorithmique: Fondements et Applications),<br />
which includes 60 permanent researchers from University Paris Diderot, CNRS, and INRIA, for a total number<br />
of members of roughly 100. Before this charge, he has been vice-Director of LRI at University Paris Sud, as<br />
well as Vice-President Research of the Computer Science Dept. of U. Paris Sud. He is currently member of the<br />
Editorial Board of Distributed Computing (DC), and Theory of Computing Systems (TOCS), among others.<br />
He is the current Program Committee Chair of the 30th Annual ACM Symposium on Principles of Distributed<br />
Computing (PODC). Before, he has acted as PC Chair for the 19th Int. Symp. on Distributed Computing<br />
(DISC 2005), and for the 13th ACM Symp. on Parallel Algorithms and Architectures (SPAA 2001). He is<br />
regular member of PC for conferences such as PODC, SPAA, DISC, ESA, ICALP, WG, MFCS, etc.<br />
Publications<br />
Number of refereed international journal: 58<br />
Number of referreed international conference with proceeding: 88<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Fraigniaud P. and Korman A.. An Optimal Ancestry Scheme and Small Universal Posets. In proc. 42th ACM<br />
Symposium on Theory of Computing (STOC), pp 611-620. (2010).<br />
• Fraigniaud P. and G. Giakkoupis. On the searchability of small-world networks with arbitrary underlying sructure.<br />
In proc. 42th ACM Symposium on Theory of Computing (STOC), pp 389-398. (2010)<br />
• Baumann H., P. Crescenzi, and Fraigniaud P.. Parsimonious Flooding in Dynamic Graphs. In proc. 28th ACM<br />
Symposium on Principles of Distributed Computing (PODC), p. 260-269. (2009)<br />
• Fraigniaud P., and G. Giakkoupis. The Effect of Power-Law Degrees on the Navigability of Small Worlds. In proc.<br />
28th ACM Symposium on Principles of Distributed Computing (PODC), p. 240-249. (2009)<br />
• A. Chaintreau, Fraigniaud P., E. Lebhar. Networks Become Navigable as Nodes Move and Forget. In proc. 35th<br />
Int. Colloquium on Automata, Languages and Programming (ICALP), LNCS 5125 , Springer, pp. 133-144. (2008)<br />
Prices, distinctions: Invited Keynote Speaker ACM PODC 2010, ICALP 2010, ICDT 2010, and ESA 2007; Best<br />
paper award ACM SPAA 2007.<br />
30
KORMAN Amos<br />
Age: 38<br />
Position: Chargé de recherches CNRS<br />
Email: amos.korman@liafa.jussieu.fr<br />
URL: www.liafa.jussieu.fr/˜pandit<br />
LIAFA, Université Paris Diderot<br />
Case 7014<br />
F-75205 Paris cedex 13<br />
Tel: (+33) 1 57 27 94 00<br />
Cursus<br />
Amos Korman received the PhD degree by Weizmann Institute of Science, Rehovot, Israel on May 8, 2006<br />
Other professional experiences:<br />
Publications<br />
Number of refereed international journal: 17<br />
Number of refereed international conference with proceeding: 26<br />
Selected publications from the past five years<br />
• Emek Y. and Korman A. and Shavitt Y..<br />
Approximating the Statistics of various Properties in Randomly Weighted Graphs.<br />
In Proc. 22nd ACM-SIAM Symp. on Discrete Algorithms (SODA), to appear. (2011).<br />
• Fraigniaud P. and Korman A.. An Optimal Ancestry Scheme and Small Universal Posets.<br />
In proc. 42th ACM Symposium on Theory of Computing (STOC), pp 611-620. (2010).<br />
• Korman A.. Labeling Schemes for Vertex Connectivity.<br />
In ACM Transactions on Algorithms (TALG), 6(2). (2010).<br />
• Emek Y. and Korman A.. Efficient Threshold Detection in a Distributed Environment.<br />
In Proc. 29th Symp. on Principles of Distributed Computing (PODC), pp 183-19. (2010).<br />
• Korman A. and Kutten S.. Distributed Verification of Minimum Spanning Trees.<br />
In Distributed Computing (DC), 20(4), pp 253-266. (2007).<br />
• Korman A.. General Compact Labeling Schemes for Dynamic Trees.<br />
In Distributed Computing (DC), 20(3), pp 179-193. (2007).<br />
Prices, distinctions:<br />
• Won the 2009 ICDCN best paper award.<br />
• Won the Dean’s Prize for Ph.D. Students in Weizmann Institute.<br />
• Won the 2005 DISC best student paper award.<br />
• Graduated at the Hebrew University with exceptional honors: Magna Cum Lauda.<br />
31
VIENNOT Laurent<br />
Age: 39<br />
Position: DR<br />
Email: laurent.viennot@inria.fr<br />
URL: gang.inria.fr/˜viennot/<br />
LIAFA, Université Paris Diderot<br />
Case 7014<br />
F-75205 Paris cedex 13<br />
Tel: (+33) 1 57 27 94 00<br />
Cursus<br />
Polytechnique, Thèse, Habilitation<br />
Other professional experiences:<br />
Laurent Viennot works on graph and network algorithms. He is a senior research scientist at french national<br />
institute on computer science INRIA since 1998. He has also been a part time teacher at Ecole Polytechnique<br />
for 7 years. He has also co-founded a startup-company on using peer-to-peer algorithms for photo sharing in<br />
2008. He is now leader of the ”GANG” INRIA project-team on network and graph algorithms.<br />
Publications<br />
Number of refereed international journal: 7<br />
Number of referreed international conference with proceeding: 32<br />
Prices, distinctions:<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Gavoille C., Godfroy Q. and Viennot L.. Multipath Spanners;. In Structural Information and Communication<br />
Complexity, 17th International Colloquium (SIROCCO), pages 211–223. (2010)<br />
• Jacquet P. and Viennot L.. Remote spanners: what to know beyond neighbors In 23rd IEEE International<br />
Parallel and Distributed Processing Symposium (IPDPS), pages 1–15. (2009)<br />
• Boufkhad Y., Mathieu F., de Montgolfier F., Perino D. and Viennot L. An upload bandwidth threshold for<br />
peer-to-peer video-on-demand scalability. In 23rd IEEE International Parallel and Distributed Processing<br />
Symposium (IPDPS), pages 1–10. (2009)<br />
• Derbel B., Gavoille C., Peleg D., and Viennot L. On the locality of distributed sparse spanner construction.<br />
In ACM Press, editor, 27th Annual ACM Symposium on Principles of Distributed Computing (PODC),<br />
pages 273–282. (2008)<br />
• Lebhar E., Fraigniaud P., and Viennot L. The inframetric model for the internet. In Proceedings of the<br />
27th IEEE International Conference on Computer Communications (INFOCOM), pages 1–9. (2008)<br />
32
Partner 2: Rennes<br />
33
BOURNEZ Olivier<br />
Age: 37<br />
Position: Professor of Computer Science at Ecole<br />
Polytechnique<br />
Email: bournez@lix.polytechnique.fr<br />
URL: http://www.lix.polytechnique.fr/˜bournez/<br />
LIX, UMR7161<br />
Ecole Polytechnique<br />
Laboratoire d’Informatique<br />
F-91128 Palaiseau Cedex<br />
Tel: (+33) 1 69 33 40 78<br />
Cursus<br />
• Since 01/09/2008, Professor of Computer Science, Ecole Polytechnique.<br />
Team Algorthmique et Complexité, lab. LIX (Ecole Polytechnique and CNRS). Director of the lab since<br />
2010.<br />
◦ 01/10/1999 – 31/08/2008, Chargé Recherche INRIA position in Nancy. Habilitation in 2006.<br />
◦ 01/11/1997 – 31/08/1998, Scientifique du Contingent, Laboratoire Verimag, Grenoble.<br />
◦ 01/09/1995 – 14/01/1999, Phd in Computer Science in 1999 in Ecole Normale Supérieure de Lyon.<br />
Thèse accessit du Prix de Thse Specif. Distinction par l’AFIT.<br />
◦ 01/09/1992 – 31/08/1996, Student of Ecole Normale Supérieure de Lyon.<br />
Research interests<br />
Computability and Complexity, in particular<br />
• Computability and Complexity with distributed models<br />
• Computability and Complexity with anonymous networks<br />
• Computability and Complexity with analog models<br />
• Algorithmic game theory.<br />
Publications<br />
Number of publications in refereed international journals: 17<br />
(Journal of Complexity, TCS, PPL, Applied Maths and Computation, Information and Computation, Fundamenta<br />
Informaticae, JLC, TOCS, JCSS, )<br />
Number of publications in referreed international conferences with proceedings: 34<br />
(ICALP, LCC, UC, MCU, RTA, TAMC, TCS, FOSSACS, HSCC, STACS )<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• G. Aupy and O. Bournez. On the number of binary-minded individuals required to compute p 1/2.<br />
Theoretical Computer Science to appear. (2010)<br />
• O. Bournez, P. Chassaing, J. Cohen, L. Gerin, and X. Koegler. On the convergence of population protocols<br />
when population goes to infinity. Applied Mathematics and Computation, 2009. (2009)<br />
• O. Bournez, M. L. Campagnolo, D. S. Graça, and E. Hainry. Polynomial differential equations compute all<br />
real computable functions on computable compact intervals. Journal of Complexity, 23(3):317 335,<br />
June 2007. (2007).<br />
• O. Bournez and M. L. Campagnolo. New Computational Paradigms. Changing Conceptions of What is Computable,<br />
chapter A Survey on Continuous Time Computations, pages 383–423. Springer-Verlag, New York,<br />
2008. (2007).<br />
• D. Barth, O. Bournez, O. Boussaton, and J. Cohen. Distributed learning of equilibria in a routing game.<br />
Parallel Processing Letters, 19:189–204, 2009. (2009).<br />
34
MOSTEFAOUI Achour<br />
Age: 41<br />
Position: Maître de Conférences<br />
Email: achour.mostefaoui@irisa.fr<br />
URL: www.irisa.fr/asap/<br />
IRISA/IFSIC, Campus de Beaulieu<br />
Avenue du Général Leclerc<br />
35042 Rennes cedex<br />
Tel: (+33) 2 99 84 71 96<br />
Cursus<br />
• Since 01/09/1996, Maîrte de Conférences.<br />
Ifsic/Irisa, Université de Rennes.<br />
◦ 01/09/1994 – 01/09/1996, ATER.<br />
Ifsic/Irisa, Université de Rennes.<br />
◦ 01/09/1991 – 01/09/1994, PhD in Computer Science.<br />
Ifsic/Irisa, Université de Rennes.<br />
Research interests<br />
Distributed algorithms and systems, in particular<br />
• Distributed data structures<br />
• Decidabiliy and efficiency in distributed computing<br />
• Fault-tolerance in distributed systems<br />
Publications<br />
Number of publications in refereed international journals: 32<br />
JACM, Siam J. Comput., Dist. Comp., IEEE TC, IEEE TPDS, IEEE TDSC, JPDC, JCSS, TCS, etc.<br />
Number of publications in referreed international conferences with proceedings: 77<br />
STOC, PODC, DISC, DSN, SPAA, ICDCS, IPDPS, OPODIS, SIROCCO, SRDS, EDCC,etc.<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Achour Mostfaoui, Michel Raynal, Corentin Travers. Narrowing power vs efficiency in synchronous set<br />
agreement: Relationship, algorithms and lower bound. Theoreical Computer Science, vol 411(1), pp. 58-<br />
69. (2010)<br />
• Achour Mostfaoui, Sergio Rajsbaum, Michel Raynal, Corentin Travers. The Combined Power of Conditions<br />
and Information on Failures to Solve Asynchronous Set Agreement. SIAM J. Comput., vol 38(4),<br />
pp. 1574-1601. (2008)<br />
• Achour Mostfaoui, Sergio Rajsbaum, Michel Raynal, Corentin Travers. On the computability power and<br />
the robustness of set agreement-oriented failure detector classes. Distributed Computing, vol 21(3),<br />
pp. 201-222 (2008)<br />
• Achour Mostfaoui, Sergio Rajsbaum, Michel Raynal. Synchronous condition-based consensus. Distributed<br />
Computing, vol 18(5), pp. 325-343 (2006)<br />
• Roy Friedman, Achour Mostfaoui, Michel Raynal. Simple and Efficient Oracle-Based Consensus Protocols for<br />
Asynchronous Byzantine Systems. IEEE Trans. Dependable Sec. Comput, vol 2(1), pp. 46-56 (2005)<br />
35
RAYNAL Michel<br />
Age: 61<br />
Position: Professeur, IUF(Membre Senior)<br />
Email: michel.raynal@irisa.fr<br />
URL: www.irisa.fr/prive/raynal/<br />
IRISA/IFSIC, Campus de Beaulieu<br />
Avenue du Général Leclerc<br />
35042 Rennes cedex<br />
Tel: (+33) 2 99 84 71 96<br />
Cursus<br />
• Since 01/12/1983, Professor.<br />
Ifsic/Irisa, Université de Rennes.<br />
• 01/11/1981 – 30/11/1983, Professor.<br />
Sup Telecom Bretagne, Brest.<br />
Research interests<br />
Distributed algorithms and systems, in particular:<br />
• Decidabiliy and efficiency in distributed computing,<br />
• Fault-tolerance and dependability, • Software transactional memory.<br />
Publications<br />
Number of publications in refereed international journals: 125<br />
( Journal of the ACM, Algorithmica, SIAM Journal of Computing, Acta Informatica, Distributed Computing,<br />
The Communications of the ACM, Information and Computation, Journal of Computer and System Sciences,<br />
JPDC, IEEE Trans. on Computers, IEEE Trans. on Software Engineering, IEEE Trans. on Knowledge and<br />
Data Engineering, IEEE Trans. on Parallel and Distributed Systems, IEEE Computer, IEEE Software, Journal<br />
of Supercomputing, Information Proc. Letters, Parallel Proc. Letters, Theoretical Computer Science, Theory of<br />
Computing Systems, Real-Time Systems Journal, The Computer Journal, etc.)<br />
Prices and distinctions: (h-index 45, Best paper award: SSS 2009, DISC 2010; Distinguihed paper Europar<br />
2010, Invited Keynote speaker: DISC 2007, AINA 2008 )<br />
Number of publications in referred int’l conferences with proceedings: 258<br />
(STOC, PODC, DISC, DSN, SPAA, ICDCS, IPDPS, OPODIS, SIROCCO, SRDS, EDCC, ICDCN, etc., etc.)<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Yehuda Afek, Eli Gafni, Sergio Rajsbaum, Michel Raynal, Corentin Travers. The k-simultaneous consensus<br />
problem. Distributed Computing, vol 22(3), pp. 185-195. (2010)<br />
• Achour Mostfaoui, Michel Raynal, Corentin Travers. Narrowing power vs efficiency in synchronous set<br />
agreement: Relationship, algorithms and lower bound. Theoretical Computer Science, vol 411(1), pp. 58-<br />
69. (2010)<br />
• Yoram Moses, Michel Raynal. Revisiting simultaneous consensus with crash failures. Journal of Parallel<br />
Distributed Computing, vol 69(4), pp. 400-409. (2009)<br />
• Achour Mostfaoui, Sergio Rajsbaum, Michel Raynal, Corentin Travers. On the computability power and<br />
the robustness of set agreement-oriented failure detector classes. Distributed Computing, vol 21(3),<br />
pp. 201-222. (2008)<br />
• Roy Friedman, Achour Mostfaoui, Michel Raynal. Simple and Efficient Oracle-Based Consensus Protocols<br />
for Asynchronous Byzantine Systems. IEEE Transactions Dependable Secure Computing, vol 2(1),<br />
pp. 46-56. (2010)<br />
36
Partner 3: Bordeaux<br />
37
GAVOILLE Cyril<br />
Age: 40<br />
Position: PR, University of Bordeaux<br />
Email: gavoille@labri.fr<br />
URL: http://www.labri.fr/perso/gavoille<br />
LaBRI, Université de Bordeaux<br />
351 cours de la Libération<br />
F-33405 Talence cedex<br />
Tel: (+33) 5 40 00 88 12<br />
Cursus<br />
• Since 2002, Professor at Bordeaux University, LaBRI<br />
◦ 1996-2002, Assistant Professor at Bordeaux University, LaBRI<br />
◦ 1993-1996, PhD at Ecole Normale Superieure de Lyon, LIP<br />
Research interests<br />
Distributed graph algorithms, distributed data-structures, distributed distance oracles, quantum information,<br />
routing algorithms.<br />
Additional information<br />
Deputy Director of the LaBRI (340 people including 140 faculties) in charge of the scientific affairs in 2009<br />
and 2010, he is currently junior member of the “Institut Universitaire de France” since 2009, a prestigious<br />
national status for five years. He is General Chair of PODC 2011 (Principles of Distributed Computing) and<br />
was Treasurer in 2010, a top world-class conference in Distributed Computing. He was also co-Chair of ICALP<br />
2010 (Int.’l Coll. on Automata, Languages and Programming), and chairman and co-chair for several workshops<br />
(SIROCCO 1999, LOCALITY at DISC 2005, AlgoTel 2003). He has participed in more than 20 international<br />
program committee conferences, in particular PODC 2005 and 2006, SPAA 2007, DISC 2007 and 2008, OPODIS<br />
2009, ESA 2011.<br />
Publications<br />
Number of publications in refereed international journals: >40<br />
(including books)<br />
Number of publications in refereed international conferences with proceedings: >90<br />
Prices, distinctions:<br />
Junior Member.<br />
h-index 27, Best Paper Award for the ACM Int.’l conference SPAA in 2006, IUF<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Fraigniaud P., Gavoille C., Ilcinkas D., Pelc A. Distributed computing with advice: Information sensitivity<br />
of graph coloring, Distributed Computing, 21(6):395-403. (2009)<br />
• Derbel B., Gavoille C., Peleg D., Viennot L. On the locality of distributed sparse spanner construction,<br />
27th Annual ACM Symposium on Principles of Distributed Computing (PODC), pages 273-282. (2008)<br />
• Gavoille C., Abraham I., Malkhi D., Nisan N., Thorup M. Compact name-independent routing with minimum<br />
stretch, ACM Transactions on Algorithms, 3(4):A37. (2008)<br />
• Gavoille C., Kosowski A., Markiewicz M. What can be observed locally? Round-based models for quantum<br />
distributed computing, 23rd International Symposium on Distributed Computing (DISC), vol. 5805 of<br />
Lecture Notes in Computer Science, pages 243-257. (2009)<br />
• Fraigniaud P., Gavoille C., Kosowski A., Lebhar E., Lotker Z. Universal Augmentation schemes for network<br />
navigability: Overcoming the √ n-barrier, 19th Annual ACM Symposium on Parallelism and Architectures<br />
(SPAA), pages 1-7. (2007)<br />
38
ILCINKAS David<br />
Age: 30<br />
Position: CNRS junior researcher (CR2)<br />
Email: david.ilcinkas@labri.fr<br />
URL: http://www.labri.fr/perso/ilcinkas<br />
LaBRI, Université de Bordeaux<br />
351 cours de la Libération<br />
F-33405 Talence cedex<br />
Tel: (+33) 5 40 00 69 12<br />
Cursus<br />
• Since 01/11/2007, Junior researcher (CR2), CNRS.<br />
Team Combinatorics and Algorithmics, lab. LaBRI (Université de Bordeaux).<br />
◦ 01/09/2006 – 31/08/2007, Post-doctoral position, Université du Québec en Outaouais & University<br />
of Ottawa & Carleton University, Canada.<br />
◦ 01/10/2003 – 31/08/2006, Allocataire-moniteur, Université Paris-Sud, team GraphComm, LRI,<br />
Orsay. PhD: in 2006.<br />
Research interests<br />
Distributed graph algorithms, in particular<br />
• Mobile agent computing<br />
• Distributed computing with an oracle / advice<br />
Additional information<br />
Program committee member of several workshops and conferences: SIROCCO 2009, DYNAS 2009 and 2010,<br />
AlgoTel 2009 and 2011, DIALM-POMC 2010, DISC 2011.<br />
Publications<br />
Number of publications in refereed international journals: 12<br />
(ACM TAlg, Algorithmica, DAM, Dist. Comp. [2], Fund. Inf., Inf. & Comp., JCSS, TCS [4])<br />
Number of publications in refereed international conferences with proceedings: 22<br />
(DISC [3], ICALP [3], IWDC, MFCS [2], OPODIS [2], PODC [2], SIROCCO [6], STACS, SWAT, WG)<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• D. Ilcinkas, D. R. Kowalski and A. Pelc. Fast Radio Broadcasting with Advice. Theoretical Computer<br />
Science, volume 411 (14-15), pages 1544-1557. (2010)<br />
• P. Fraigniaud, D. Ilcinkas and A. Pelc. Communication algorithms with advice. Journal of Computer and<br />
System Sciences, volume 76 (3-4), pages 222-232. (2010)<br />
• P. Fraigniaud, C. Gavoille, D. Ilcinkas and A. Pelc. Distributed computing with advice: information sensitivity<br />
of graph coloring. Dist. Comp., volume 21 (6), pages 395-403. (2009)<br />
• R. Cohen, P. Fraigniaud, D. Ilcinkas, A. Korman and D. Peleg. Label-Guided Graph Exploration by a Finite<br />
Automaton. ACM Trans. on Algorithms, volume 4 (4), article 42. (2008)<br />
• P. Fraigniaud, D. Ilcinkas and A. Pelc. Tree Exploration with Advice. Information and Computation, volume<br />
206 (11), pages 1276-1287. (2008)<br />
————————————————–<br />
39
JOHNEN Colette<br />
Age: 49<br />
Position: Full Professor (PR2)<br />
Email: colette.johnen@labri.fr<br />
URL: http://www.labri.fr/∼johnen<br />
LaBRI, Université de Bordeaux<br />
351 cours de la Libération<br />
F-33405 Talence cedex<br />
Tel: (+33) 5 40 00 60 47<br />
Cursus<br />
• Since 01/10/2008, Full Professor, Université de Bordeaux<br />
Team Combinatorics and Algorithmics, lab. LaBRI (CNRS - UMR 5800).<br />
◦ 01/12/1988 – 31/08/2008, Associated Professor Université de Paris-Sud<br />
Team Parallelism, lab. LRI ( CNRS - UMR 8623).<br />
Research interests<br />
Fault-Tolerant Distributed algorithms<br />
Additional information<br />
PhD Students since 2006 : 3<br />
2004 - 2007 Advisor for 80% of Le Huy Nguyen’s PhD thesis (Dr. Pierre Fraigniaud was the co-advisor).<br />
2005 - 2009 Advisor for 20% of Florent Kaisser’s PhD thesis (Pr. Véronique Vèque was the co-advisor).<br />
2008 - 2011 Advisor for 80% of Fouzi Mekhaldi’s PhD thesis (Pr. Véronique Vèque was the co-advisor).<br />
Publications<br />
Number of publications in refereed international journals : 7<br />
(TAAS, TCS, Distributed Computing [2], IPL, PPL, JPDC )<br />
Number of publications in refereed international conferences with proceedings: 34<br />
(AlgoSensors, ICDCN, ICDCS, IPDPS, EuroPar, SSS, SRDS, OPODIS, DISC, PODC, ...)<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Colette Johnen and Fouzi Mekhaldi. Robust Self-stabilizing Construction of Bounded Size Weight-Based<br />
Clusters. Euro-Par, LNCS 6271, pages 535-546. (2010)<br />
• Colette Johnen and Lisa Higham. Fault-Tolerant Implementations of Regular Registers by Safe Registers<br />
with Applications to Networks. ICDCN, LNCS 5408, pages 337-348. (2009)<br />
• Kajari Ghosh Dastidar, Ted Herman and Colette Johnen. Safe peer-to-peer self-downloading. TAAS, vol 3<br />
(4), paper 19. (2008)<br />
• Joffroy Beauquier, Maria Gradinariu and Colette Johnen: Randomized self-stabilizing and space optimal<br />
leader election under arbitrary scheduler on rings. Distributed Computing, vol 20 (1), pages 75-93. (2007)<br />
• Joffroy Beauquier, Colette Johnen, Stéphane Messika. All k -Bounded Policies Are Equivalent for Selfstabilization.<br />
SSS LNCS 4280, pages 82-94. (2006)<br />
40
Age: 47<br />
Position: CNRS senior researcher<br />
(DR2)<br />
Email: klasing@labri.fr<br />
URL: http://www.labri.fr/perso/klasing<br />
KLASING Ralf<br />
LaBRI, Université de Bordeaux<br />
351 cours de la Libération<br />
F-33405 Talence cedex<br />
Tel: (+33) 5 40 00 35 24<br />
Cursus<br />
Ralf Klasing received the PhD degree from the University of Paderborn in 1995. From 1995 to 1997, he was an<br />
Assistant Professor at the University of Kiel. From 1997 to 1998, he was a Research Fellow at the University<br />
of Warwick. From 1998 to 2000, he was an Assistant Professor at RWTH Aachen. From 2000 to 2002, he was<br />
a Lecturer at King’s College London. In 2002, he joined the CNRS as a permanent researcher. From 2002 to<br />
2005, he was affiliated to the laboratory I3S in Sophia Antipolis. Currently, he is affiliated to the laboratory<br />
LaBRI in Bordeaux. In 2009, he received the HDR degree from the University Bordeaux 1. In 2010, he was<br />
promoted to Senior Researcher (DR2 CNRS).<br />
Research interests<br />
Design and Analysis of Algorithms, Communication algorithms in networks, Approximation algorithms for<br />
combinatorially hard problems, Algorithmic methods for telecommunication, Distributed algorithms.<br />
Additional information<br />
Head of the Combinatorics and Algorithms team of the LaBRI. Member of the Editorial Boards of Theoretical<br />
Computer Science, Discrete Applied Mathematics, Wireless Networks, Networks, Journal of Interconnection<br />
Networks, Parallel Processing Letters, Algorithmic Operations Research, Fundamenta Informaticae, Computing<br />
and Informatics. Member of the Program Committees of SIROCCO 2006, SIROCCO 2009, MFCS 2009,<br />
ADHOC-NOW 2009, OPODIS 2009, STACS 2010, ALGOSENSORS 2010, ADHOC-NOW 2010, IWOCA 2010,<br />
SIROCCO 2011, FOMC 2011, ADHOC-NOW 2011, IWOCA 2011. Member of the Evaluation Committee of<br />
the programme ANR Défis 2009. Workshops Chair of ICALP 2010.<br />
Publications<br />
Number of books: 2 (Springer Monographs)<br />
Number of book chapters: 3<br />
Number of publications in refereed international journals: 33 (SIAM J. Discr. Math., DAM [2],<br />
Inf. & Comp. [2], JCSS, TCS [8], IEEE Trans. Par. & Distr. Syst., J. Par. & Distr. Comp., Algorithmica,<br />
Networks [4], R.A.I.R.O., ...)<br />
Number of publications in refereed international conferences with proceedings: 35<br />
(ICALP, STACS [4], ESA, MFCS, SWAT, WG [2], DISC [2], OPODIS [4], SIROCCO [4], CIAC [2], ISAAC<br />
[2], ...)<br />
Selected publications from the past five years<br />
• L. Gasieniec, R. Klasing, R. Martin, A. Navarra, X. Zhang. Fast Periodic Graph Exploration with Constant<br />
Memory. Journal of Computer and System Sciences 74, No. 5, 808–822. (2008)<br />
• J. Hromkovič, P. Kanarek, R. Klasing, K. Lorys, W. Unger, H. Wagener. On the Size of Permutation Networks<br />
and Consequences for Efficient Simulation of Hypercube Algorithms on Bounded-Degree<br />
Networks. SIAM Journal on Discrete Mathematics 23, No. 3, 1612–1645. (2009)<br />
• C. Cooper, D. Ilcinkas, R. Klasing, A. Kosowski. Derandomizing Random Walks in Undirected Graphs<br />
Using Locally Fair Exploration Strategies. In: Proc. 36th International Colloquium on Automata, Languages<br />
and Programming (ICALP 2009 ), Lecture Notes in Computer Science 5556, Springer-Verlag, 411–422. (2009)<br />
• C. Cooper, R. Klasing and T. Radzik. Locating and repairing faults in a network with mobile agents.<br />
Theoretical Computer Science 411(14–15):1638–1647. (2010)<br />
• R. Klasing, A. Kosowski and A. Navarra. Taking Advantage of Symmetries: Gathering of many Asynchronous<br />
Oblivious Robots on a Ring. Theoretical Computer Science 411(34–36):3235–3246. (2010)<br />
41
KOSOWSKI Adrian<br />
Age: 24<br />
Position: INRIA researcher (CR1)<br />
Email: adrian.kosowski@labri.fr<br />
URL: http://www.labri.fr/perso/kosowski<br />
INRIA Bordeaux Sud-Ouest, LaBRI<br />
351 cours de la Libération<br />
F-33405 Talence cedex<br />
Tel: (+33) 5 40 00 69 12<br />
Cursus<br />
• Since 15/10/2010, Researcher (CR1), INRIA.<br />
CEPAGE project, lab. LaBRI (INRIA Bordeaux Sud-Ouest).<br />
◦ 01/10/2007 – 31/10/2010, Assistant Professor, Department of Algorithms and System Modeling,<br />
Gdańsk University of Technology, Poland.<br />
◦ 01/07/2008 – 30/06/2009, Post-doctoral position, Team Combinatorics and Algorithmics, lab.<br />
LaBRI (Université de Bordeaux).<br />
◦ 01/10/2005 – 30/09/2007, Research and teaching assistant, Department of Algorithms and<br />
System Modeling, Gdańsk University of Technology, Poland. PhD: in 2007.<br />
Research interests<br />
• Distributed algorithms, in particular: distributed graph coloring algorithms, mobile agent computing,<br />
self-stabilization and self-organization.<br />
• Combinatorial optimization algorithms, in particular: packing and routing paths in graphs, graph<br />
coloring, graph exploration, computational geometry.<br />
• Quantum information and distributed quantum computing.<br />
Publications<br />
Number of publications in refereed international journals: 19<br />
(Wireless Networks, Theor. Comp. Sci. [3], Disc. App. Math. [3], Disc. Math. [2], Networks [2], Inf.<br />
Process. Lett. [2], Comp. Geom. Theory, and others)<br />
Number of publications in refereed international conferences with proceedings: 30<br />
(PODC [2], ICALP, SPAA, MFCS [2], DISC [3], ISAAC, SPIRE, SIROCCO [2], OPODIS [4], Euro-Par,<br />
and others)<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• R. Klasing, A. Kosowski, A. Navarra. Taking advantage of symmetries: Gathering of many asynchronous<br />
oblivious robots on a ring. Theoretical Computer Science, vol 411 (34-36), pp 3235-3246. (2010)<br />
• C. Gavoille, A. Kosowski, M. Markiewicz. What can be observed locally? Round-based models for<br />
quantum distributed computing. Proceedings of 23rd International Symposium on Distributed Computing,<br />
Lecture Notes in Computer Science, vol 5805, pp 243-257. (2009)<br />
• C. Gavoille, R. Klasing, A. Kosowski, L. Kuszner, A. Navarra. On the complexity of distributed graph<br />
coloring with local minimality constraints. Networks, vol 54 (1), pp 12-19. (2009)<br />
• P. Fraigniaud, C. Gavoille, A. Kosowski, E. Lebhar, Z. Lotker. Universal augmentation schemes for<br />
network navigability. Theoretical Computer Science, vol 410 (21-23), pp 1970-1981. (2009)<br />
• A. Kosowski. The maximum edge-disjoint paths problem in complete graphs. Theoretical Computer<br />
Science, vol 399 (1-2), pp 128-140. (2008)<br />
42
Age: 30<br />
Position: Associate professor (Maître de<br />
conférence)<br />
Email: travers@labri.fr<br />
URL: www.labri.fr/perso/travers<br />
TRAVERS Corentin<br />
LaBRI, Université de Bordeaux<br />
351 cours de la Libération<br />
F-33405 Talence cedex<br />
Tel: (+33) 5 40 00 35 29<br />
Cursus<br />
• Since 01/09/2010, Assistant Professor (Maître de conférence), ENSEIRB.<br />
Team Combinatorics and Algorithmics, lab. LaBRI (Université de Bordeaux).<br />
◦ 01/10/2009 – 31/08/2010, Post-doctoral position, Université Pierre et Marie Curie.<br />
◦ 01/10/2008 – 30/09/2009, Post-doctoral position, The Technion, Israel.<br />
◦ 01/01/2008 – 30/09/2008, Post-doctoral position, Universidad Politécnica de Madrid.<br />
◦ 01/10/2004 – 31/12/2007, Research/Teaching assistant, Université de Rennes 1, IRISA. PhD: in<br />
2007.<br />
Research interests<br />
Theory of distributed Computing, in particular<br />
• Fault-tolerant computing<br />
• Distributed computing with oracles / advice<br />
Additional information<br />
Program committee member of the conference and the workshop: ICDCS 2008 and WRAS 2010.<br />
Won the best paper award of the conference ICDCN 2011 for the paper Generating Fast Indulgent Algorithms<br />
coauthored with D. Alistarh, S. Gilbert and R. Guerraoui.<br />
Publications<br />
Number of publications in refereed international journals: 10<br />
( Dist. Comp. [2], Parallel Processing Letters, Trans. Parallel Distrib. Syst., J. Parallel Distrib. Comput.,<br />
SIAM J. Comput., Inf. Process. Lett., Theor. Comput. Sci. [2], Theory Comput. Syst.)<br />
Number of publications in refereed international conferences with proceedings: 21<br />
(COCOON, DISC [3], ICDCN [3], ISAAC, LATIN, OPODIS [3], PODC [2], PRDC [3], SIROCCO [2], SRDS<br />
[2])<br />
Selected publications from the past five years<br />
(Authors are given by the alphabetical order of their name)<br />
• Y. Afek, E. Gafni, S. Rajsbaum, M. Raynal, C. Travers. The k-simultaneous Consensus Problem. Distributed<br />
Computing, vol 22(3), pp 185-195. (2010)<br />
• A. Mostfaoui, M. Raynal, C. Travers. Narrowing Power vs Efficiency in Synchronous Set Agreement:<br />
Relationship, Algorithms and Lower Bound. Theoretical Computer Science, vol 411 (1), pp 58-69. (2010)<br />
• E. Gafni, A. Mostfaoui, M. Raynal, C. Travers. From Adaptive Renaming to Set Agreement. Theoretical<br />
Computer Science, vol 410(14), pp 1328-1335. (2009)<br />
• A. Mostfaoui, S. Rajsbaum, M. Raynal, C. Travers. On the Computability Power and the Robustness of<br />
Set Agreement-oriented Failure Detector Classes. Distributed Computing, vol 21(3), pp 201-222. (2008)<br />
• A. Mostéfaoui, S. Rajsbaum, M. Raynal, C. Travers. The Combined Power of Conditions and Information<br />
on Failures to Solve Asynchronous Set Agreement. SIAM Journal of Computing, vol 38(4), pp1574-<br />
1601. (2008)<br />
43
7.3 Staff involvment in other contracts<br />
Partner 1: Paris<br />
Part. Name PM(to<br />
do)<br />
1 Delporte 2<br />
1 Fauconnier 7<br />
1 Fraigniaud 1<br />
1 Viennot 9,6<br />
1 Fraigniaud 6<br />
1 Korman 3<br />
1 Viennot 3<br />
Project name, financing,<br />
institution, grant<br />
allocated<br />
ANR VERSO (105 000<br />
E)<br />
ANR VERSO (104 000<br />
E)<br />
Project<br />
title<br />
Coordinator<br />
name<br />
SHAMAN Tixeuil 2008-2012<br />
Start and end dates<br />
PROSE Chaintreau Sept 2009-Aug 2012<br />
Partner 2: Rennes<br />
Part. Name PM Project name, financing,<br />
institution, grant<br />
allocated<br />
2 Bournez 19<br />
2 Kermarrec 5<br />
2 Mostefaoui 19<br />
2 Raynal 9<br />
2 Mostefaoui 8<br />
2 Raynal 8<br />
2 Kermarrec 48<br />
ANR VERSO (105 000<br />
E)<br />
ANR ARPÈGE (99<br />
008 E)<br />
ERC Junior - EC (1<br />
000 000 E)<br />
Project<br />
title<br />
Coordinator<br />
name<br />
SHAMAN Tixeuil 2008-2012<br />
STREAMS Oster 2010-2013<br />
GOOSPLE Kermarrec 2008-2013<br />
Start and end dates<br />
44
Partner 3: Bordeaux<br />
Part. Name PM Project name, financing<br />
institution, grant<br />
allocated<br />
3 Adrian 1 Royal Society grant,<br />
Kosowski<br />
12000 pounds<br />
3 Ralf Klasing<br />
3 David<br />
Ilcinkas<br />
3 David<br />
Ilcinkas<br />
3 Cyril<br />
Gavoille<br />
2 Royal Society grant,<br />
12000 pounds<br />
1 Royal Society grant,<br />
12000 pounds<br />
Leszek Gasieniec<br />
4 FP7 STREP Project,<br />
European Commission,<br />
3 150 000 euros<br />
Dimitri Papadimitriou<br />
1 FP7 STREP Project,<br />
European Commission,<br />
3 150 000 euros<br />
Project<br />
title<br />
Routing<br />
EULER<br />
- Experimental<br />
Update-<br />
Less<br />
Evolutive<br />
Routing<br />
Coordinator<br />
name<br />
Leszek Gasieniec<br />
Leszek Gasieniec<br />
SERENE<br />
- SEarch,<br />
RENdezvous,<br />
and<br />
Explore<br />
SERENE<br />
- SEarch,<br />
RENdezvous,<br />
and<br />
Explore<br />
SERENE<br />
- SEarch,<br />
RENdezvous,<br />
and<br />
Explore<br />
EULER<br />
- Experimental<br />
Update-<br />
Less<br />
Evolutive<br />
Dimitri Papadimitriou<br />
Start and end dates<br />
Jan. 1, 2011 - Dec.<br />
31, 2012<br />
Jan. 1, 2011 - Dec.<br />
31, 2012<br />
Jan. 1, 2011 - Dec.<br />
31, 2012<br />
Oct. 1, 2010 - Sept.<br />
30, 2013<br />
Oct. 1, 2010 - Sept.<br />
30, 2013<br />
45